Connect with us

AI

Building a medical image search platform on AWS

Improving radiologist efficiency and preventing burnout is a primary goal for healthcare providers. A nationwide study published in Mayo Clinic Proceedings in 2015 showed radiologist burnout percentage at a concerning 61% [1]. In additon, the report concludes that “burnout and satisfaction with work-life balance in US physicians worsened from 2011 to 2014. More than half […]

Published

on

Improving radiologist efficiency and preventing burnout is a primary goal for healthcare providers. A nationwide study published in Mayo Clinic Proceedings in 2015 showed radiologist burnout percentage at a concerning 61% [1]. In additon, the report concludes that “burnout and satisfaction with work-life balance in US physicians worsened from 2011 to 2014. More than half of US physicians are now experiencing professional burnout.”[2] As technologists, we’re looking for ways to put new and innovative solutions in the hands of physicians to make them more efficient, reduce burnout, and improve care quality.

To reduce burnout and improve value-based care through data-driven decision-making, Artificial Intelligence (AI) can be used to unlock the information trapped in the vast amount of unstructured data (e.g. images, texts, and voice) and create clinically actionable knowledge base. AWS AI services can derive insights and relationships from free-form medical reports, automate the knowledge sharing process, and eventually improve personalized care experience.

In this post, we use Convolutional Neural Networks (CNN) as a feature extractor to convert medical images into a one-dimensional feature vector with a size of 1024. We call this process medical image embedding. Then we index the image feature vector using the K-nearest neighbors (KNN) algorithm in Amazon Elasticsearch Service (Amazon ES) to build a similarity-based image retrieval system. Additionally, we use the AWS managed natural language processing (NLP) service Amazon Comprehend Medical to perform named entity recognition (NER) against free text clinical reports. The detected named entities are also linked to medical ontology, ICD-10-CM, to enable simple aggregation and distribution analysis. The presented solution also includes a front-end React web application and backend GraphQL API managed by AWS Amplify and AWS AppSync, and authentication is handled by Amazon Cognito.

After deploying this working solution, the end-users (healthcare providers) can search through a repository of unstructured free text and medical images, conduct analytical operations, and use it in medical training and clinical decision support. This eliminates the need to manually analyze all the images and reports and get to the most relevant ones. Using a system like this improves the provider’s efficiency. The following graphic shows an example end result of the deployed application.

Dataset and architecture

We use the MIMIC CXR dataset to demonstrate how this working solution can benefit healthcare providers, in particular, radiologists. MIMIC CXR is a publicly available database of chest X-ray images in DICOM format and the associated radiology reports as free text files[3]. The methods for data collection and the data structures in this dataset have been well documented and are very detailed [3]. Also, this is a restricted-access resource. To access the files, you must be a registered user and sign the data use agreement. The following sections provide more details on the components of the architecture.

The following diagram illustrates the solution architecture.

The architecture is comprised of the offline data transformation and online query components. The offline data transformation step, the unstructured data, including free texts and image files, is converted into structured data.

Electronic Heath Record (EHR) radiology reports as free text are processed using Amazon Comprehend Medical, an NLP service that uses machine learning to extract relevant medical information from unstructured text, such as medical conditions including clinical signs, diagnosis, and symptoms. The named entities are identified and mapped to structured vocabularies, such as ICD-10 Clinical Modifications (CMs) ontology. The unstructured text plus structured named entities are stored in Amazon ES to enable free text search and term aggregations.

The medical images from Picture Archiving and Communication System (PACS) are converted into vector representations using a pretrained deep learning model deployed in an Amazon Elastic Container Service (Amazon ECS) AWS Fargate cluster. Similar visual search on AWS has been published previously for online retail product image search. It used an Amazon SageMaker built-in KNN algorithm for similarity search, which supports different index types and distance metrics.

We took advantage of the KNN for Amazon ES to find the k closest images from a feature space as demonstrated on the GitHub repo. KNN search is supported in Amazon ES version 7.4+. The container running on the ECS Fargate cluster reads medical images in DICOM format, carries out image embedding using a pretrained model, and saves a PNG thumbnail in an Amazon Simple Storage Service (Amazon S3) bucket, which serves as the storage for AWS Amplify React web application. It also parses out the DICOM image metadata and saves them in Amazon DynamoDB. The image vectors are saved in an Elasticsearch cluster and are used for the KNN visual search, which is implemented in an AWS Lambda function.

The unstructured data from EHR and PACS needs to be transferred to Amazon S3 to trigger the serverless data processing pipeline through the Lambda functions. You can achieve this data transfer by using AWS Storage Gateway or AWS DataSync, which is out of the scope of this post. The online query API, including the GraphQL schemas and resolvers, was developed in AWS AppSync. The front-end web application was developed using the Amplify React framework, which can be deployed using the Amplify CLI. The detailed AWS CloudFormation templates and sample code are available in the Github repo.

Solution overview

To deploy the solution, you complete the following steps:

  1. Deploy the Amplify React web application for online search.
  2. Deploy the image-embedding container to AWS Fargate.
  3. Deploy the data-processing pipeline and AWS AppSync API.

Deploying the Amplify React web application

The first step creates the Amplify React web application, as shown in the following diagram.

  1. Install and configure the AWS Command Line Interface (AWS CLI).
  2. Install the AWS Amplify CLI.
  3. Clone the code base with stepwise instructions.
  4. Go to your code base folder and initialize the Amplify app using the command amplify init. You must answer a series of questions, like the name of the Amplify app.

After this step, you have the following changes in your local and cloud environments:

  • A new folder named amplify is created in your local environment
  • A file named aws-exports.js is created in local the src folder
  • A new Amplify app is created on the AWS Cloud with the name provided during deployment (for example, medical-image-search)
  • A CloudFormation stack is created on the AWS Cloud with the prefix amplify-<AppName>

You create authentication and storage services for your Amplify app afterwards using the following commands:

amplify add auth
amplify add storage
amplify push

When the CloudFormation nested stacks for authentication and storage are successfully deployed, you can see the new Amazon Cognito user pool as the authentication backend and S3 bucket as the storage backend are created. Save the Amazon Cognito user pool ID and S3 bucket name from the Outputs tab of the corresponding CloudFormation nested stack (you use these later).

The following screenshot shows the location of the user pool ID on the Outputs tab.

The following screenshot shows the location of the bucket name on the Outputs tab.

Deploying the image-embedding container to AWS Fargate

We use the Amazon SageMaker Inference Toolkit to serve the PyTorch inference model, which converts a medical image in DICOM format into a feature vector with the size of 1024. To create a container with all the dependencies, you can either use pre-built deep learning container images or derive a Dockerfile from the Amazon Sagemaker Pytorch inference CPU container, like the one from the GitHub repo, in the container folder. You can build the Docker container and push it to Amazon ECR manually or by running the shell script build_and_push.sh. You use the repository image URI for the Docker container later to deploy the AWS Fargate cluster.

The following screenshot shows the sagemaker-pytorch-inference repository on the Amazon ECR console.

We use Multi Model Server (MMS) to serve the inference endpoint. You need to install MMS with pip locally, use the Model archiver CLI to package model artifacts into a single model archive .mar file, and upload it to an S3 bucket to be served by a containerized inference endpoint. The model inference handler is defined in dicom_featurization_service.py in the MMS folder. If you have a domain-specific pretrained Pytorch model, place the model.pth file in the MMS folder; otherwise, the handler uses a pretrained DenseNET121[4] for image processing. See the following code:

model_file_path = os.path.join(model_dir, "model.pth")
if os.path.isfile(model_file_path): model = torch.load(model_file_path) else: model = models.densenet121(pretrained=True) model = model._modules.get('features') model.add_module("end_relu", nn.ReLU()) model.add_module("end_globpool", nn.AdaptiveAvgPool2d((1, 1))) model.add_module("end_flatten", nn.Flatten())
model = model.to(self.device)
model.eval()

The intermediate results of this CNN-based model is to represent images as feature vectors. In other words, the convolutional layers before the final classification layer is flattened to convert feature layers to a vector representation. Run the following command in the MMS folder to package up the model archive file:

model-archiver -f --model-name dicom_featurization_service --model-path ./ --handler dicom_featurization_service:handle --export-path ./

The preceding code generates a package file named dicom_featurization_service.mar. Create a new S3 bucket and upload the package file to that bucket with public read Access Control List (ACL). See the following code:

aws s3 cp ./dicom_featurization_service.mar s3://<S3bucketname>/ --acl public-read --profile <profilename>

You’re now ready to deploy the image-embedding inference model to the AWS Fargate cluster using the CloudFormation template ecsfargate.yaml in the CloudFormationTemplates folder. You can deploy using the AWS CLI: go to the CloudFormationTemplates folder and copy the following command:

aws cloudformation deploy --capabilities CAPABILITY_IAM --template-file ./ecsfargate.yaml --stack-name <stackname> --parameter-overrides ImageUrl=<imageURI> InferenceModelS3Location=https://<S3bucketname>.s3.amazonaws.com/dicom_featurization_service.mar --profile <profilename>

You need to replace the following placeholders:

  • stackname – A unique name to refer to this CloudFormation stack
  • imageURI – The image URI for the MMS Docker container uploaded in Amazon ECR
  • S3bucketname – The MMS package in the S3 bucket, such as https://<S3bucketname>.s3.amazonaws.com/dicom_featurization_service.mar
  • profilename – Your AWS CLI profile name (default if not named)

Alternatively, you can choose Launch stack for the following Regions:

  • us-east-1

  • us-west-2

After the CloudFormation stack creation is complete, go to the stack Outputs tab on the AWS CloudFormation console and copy the InferenceAPIUrl for later deployment. See the following screenshot.

You can delete this stack after the offline image embedding jobs are finished to save costs, because it’s not used for online queries.

Deploying the data-processing pipeline and AWS AppSync API

You deploy the image and free text data-processing pipeline and AWS AppSync API backend through another CloudFormation template named AppSyncBackend.yaml in the CloudFormationTemplates folder, which creates the AWS resources for this solution. See the following solution architecture.

To deploy this stack using the AWS CLI, go to the CloudFormationTemplates folder and copy the following command:

aws cloudformation deploy --capabilities CAPABILITY_NAMED_IAM --template-file ./AppSyncBackend.yaml --stack-name <stackname> --parameter-overrides AuthorizationUserPool=<CFN_output_auth> PNGBucketName=<CFN_output_storage> InferenceEndpointURL=<inferenceAPIUrl> --profile <profilename>

Replace the following placeholders:

  • stackname – A unique name to refer to this CloudFormation stack
  • AuthorizationUserPool – Amazon Cognito user pool
  • PNGBucketName – Amazon S3 bucket name
  • InferenceEndpointURL – The inference API endpoint
  • Profilename – The AWS CLI profile name (use default if not named)

Alternatively, you can choose Launch stack for the following Regions:

  • us-east-1

  • us-west-2

You can download the Lambda function for medical image processing, CMprocessLambdaFunction.py, and its dependency layer separately if you deploy this stack in AWS Regions other than us-east-1 and us-west-2. Because their file size exceeds the CloudFormation template limit, you need to upload them to your own S3 bucket (either create a new S3 bucket or use the existing one, like the aforementioned S3 bucket for hosting the MMS model package file) and override the LambdaBucket mapping parameter using your own bucket name.

Save the AWS AppySync API URL and AWS Region from the settings on the AWS AppSync console.

Edit the src/aws-exports.js file in your local environment and replace the placeholders with those values:

const awsmobile = { "aws_appsync_graphqlEndpoint": "<AppSync API URL>", "aws_appsync_region": "<AWS AppSync Region>", "aws_appsync_authenticationType": "AMAZON_COGNITO_USER_POOLS"
};

After this stack is successfully deployed, you’re ready to use this solution. If you have in-house EHR and PACS databases, you can set up the AWS Storage Gateway to transfer data to the S3 bucket to trigger the transformation jobs.

Alternatively, you can use the public dataset MIMIC CXR: download the MIMIC CXR dataset from PhysioNet (to access the files, you must be a credentialed user and sign the data use agreement for the project) and upload the DICOM files to the S3 bucket mimic-cxr-dicom- and the free text radiology report to the S3 bucket mimic-cxr-report-. If everything works as expected, you should see the new records created in the DynamoDB table medical-image-metadata and the Amazon ES domain medical-image-search.

You can test the Amplify React web application locally by running the following command:

npm install && npm start

Or you can publish the React web app by deploying it in Amazon S3 with AWS CloudFront distribution, by first entering the following code:

amplify hosting add

Then, enter the following code:

amplify publish

You can see the hosting endpoint for the Amplify React web application after deployment.

Conclusion

We have demonstrated how to deploy, index and search medical images on AWS, which segregates the offline data ingestion and online search query functions. You can use AWS AI services to transform unstructured data, for example the medical images and radiology reports, into structured ones.

By default, the solution uses a general-purpose model trained on ImageNET to extract features from images. However, this default model may not be accurate enough to extract medical image features because there are fundamental differences in appearance, size, and features between medical images in its raw form. Such differences make it hard to train commonly adopted triplet-based learning networks [5], where semantically relevant images or objects can be easily defined or ranked.

To improve search relevancy, we performed an experiment by using the same MIMIC CXR dataset and the derived diagnosis labels to train a weakly supervised disease classification network similar to Wang et. Al [6]. We found this domain-specific pretrained model yielded qualitatively better visual search results. So it’s recommended to bring your own model (BYOM) to this search platform for real-world implementation.

The methods presented here enable you to perform indexing, searching and aggregation against unstructured images in addition to free text. It sets the stage for future work that can combine these features for multimodal medical image search engine. Information retrieval from unstructured corpuses of clinical notes and images is a time-consuming and tedious task. Our solution allows radiologists to become more efficient and help them reduce potential burnout.

To find the latest development to this solution, check out medical image search on GitHub.

Reference:

  1. https://www.radiologybusiness.com/topics/leadership/radiologist-burnout-are-we-done-yet
  2. https://www.mayoclinicproceedings.org/article/S0025-6196(15)00716-8/abstract#secsectitle0010
  3. Johnson, Alistair EW, et al. “MIMIC-CXR, a de-identified publicly available database of chest radiographs with free-text reports.” Scientific Data 6, 2019.
  4. Huang, Gao, et al. “Densely connected convolutional networks.” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017.
  5. Wang, Jiang, et al. “Learning fine-grained image similarity with deep ranking.” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2014.
  6. Wang, Xiaosong, et al. “Chestx-ray8: Hospital-scale chest x-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases.” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017.

About the Authors

 Gang Fu is a Healthcare Solution Architect at AWS. He holds a PhD in Pharmaceutical Science from the University of Mississippi and has over ten years of technology and biomedical research experience. He is passionate about technology and the impact it can make on healthcare.

Ujjwal Ratan is a Principal Machine Learning Specialist Solution Architect in the Global Healthcare and Lifesciences team at Amazon Web Services. He works on the application of machine learning and deep learning to real world industry problems like medical imaging, unstructured clinical text, genomics, precision medicine, clinical trials and quality of care improvement. He has expertise in scaling machine learning/deep learning algorithms on the AWS cloud for accelerated training and inference. In his free time, he enjoys listening to (and playing) music and taking unplanned road trips with his family.

Erhan Bas is a Senior Applied Scientist in the AWS Rekognition team, currently developing deep learning algorithms for computer vision applications. His expertise is in machine learning and large scale image analysis techniques, especially in biomedical, life sciences and industrial inspection technologies. He enjoys playing video games, drinking coffee, and traveling with his family.

Source: https://aws.amazon.com/blogs/machine-learning/building-a-medical-image-search-platform-on-aws/

AI

5 Work From Home Office Essentials

Working remotely from home had been increasing in popularity, but it’s now become a necessity for many professionals due to the pandemic. “Some companies are eager to reopen their doors and return to the office, but a large number of employer and employees are making the transitional work environment a permanent change.”  They can’t guarantee […]

The post 5 Work From Home Office Essentials appeared first on Aiiot Talk – Artificial Intelligence | Internet of Things | Technology.

Published

on

Working remotely from home had been increasing in popularity, but it’s now become a necessity for many professionals due to the pandemic.

“Some companies are eager to reopen their doors and return to the office, but a large number of employer and employees are making the transitional work environment a permanent change.” 

They can’t guarantee their health and safety in a socially-crowded space, plus, companies are able to save tons of money they would have spent on their commercial lease or mortgage payments.

That’s not the say that working from home doesn’t come at its own costs, however. It can lead to a huge hit in productivity without the right equipment in place. To maximize your performance and efficiency in a remote setting, be sure to purchase these five office essentials.

1. Powerful PC

This one probably feels like an obvious pointer, but let’s knock it off our list. You won’t be able to get by with a make-shift work station and in today’s digital domain, your computer will be at the core of everything you do.

Never-ending loading wheels, delayed downloads, and slow rendering will add seconds to every task you do, so if your company didn’t provide you with a workhorse computer tower, you might look into investing in one yourself, then deduct the cost in your tax return.

Depending on your line of work, it might make more sense to go for a laptop vs a desktop computer. Unless your tasks demand super sophisticated software and large storage space, you can probably get by with a portable PC. That way, when coffee shops begin to reopen and allow patrons to sit inside, you can work on-the-go without feeling tethered to your desk.

2. Ergonomic Office Chair

If you’re looking at a long-term remote situation, it’s worth spending the big bucks on an ergonomic office chair. You should feel comfortably locked into your seat for eight hours a day—at least if you want to concentrate on your workflow, rather than the cramp in your back.

Shop around for an office chair that’s sophisticated in design and specifically built to hold the human body. Some stand-out features you should look out for include:

  • Targeted support around the lumbar spine
  • Adjustable height so you can adjust the seat as necessary for your arms to rest naturally on the keyboard
  • Swivel base to effortlessly turn your body, preventing neck strain
  • Cushioned seat to comfort your tailbone
  • Ventilated fabric that promotes airflow so you don’t feel overheated when sitting in the chair for several hours

You might have to pay a couple of hundred dollars for the best-of-the-line features, but there is another item that might qualify as an eligible tax deduction—just be sure to keep all your receipts organized with a document scanner in case the IRS raises their eyebrows and issues an audit.

3. Wireless Keyboard

If you want to type faster and feel better while you’re at it, then a wireless keyboard is clutch. They enable you to bring the keys closer, decreasing the extension length of your arms and accompanying shoulder strain.

“It also helps reduce the strain on your eyes by moving the bright screen farther away from your direct line of sight.” 

And, last but not least, the keys are placed in an ergonomic position for a more natural finger splay, with ample cushioning wrist cushioning that helps prevent overuse injuries such as a carpal tunnel.

4. Noise-cancelling Headphones

To truly get in the zone, you should block out distractions with headphones the cancel noise in your environment—especially if your work station is set up in a common area. Other tips to stay focused include installing a website blocker and leaving your cellphone on the other side of the room.

5. House plant or flowers

People are scientifically proven to be more productive when working near fresh flowers or lush greenery. The good news is that you don’t need to have a green thumb or natural lighting to achieve this effect—even artificial foliage can brighten your mood and improve your performance.

Working from home sometimes can feel like you’re locked inside all day, so bringing the outside world inside your space can help ward off burnout.

Take these tips with you into 2021 and set yourself up for success in your new home office setting.

Source: https://www.aiiottalk.com/business/work-from-home-office-essentials/

Continue Reading

AI

zomato digitizes menus using Amazon Textract and Amazon SageMaker

This post is co-written by Chiranjeev Ghai, ML Engineer at zomato. zomato is a global food-tech company based in India. Are you the kind of person who has very specific cravings? Maybe when the mood hits, you don’t want just any kind of Indian food—you want Chicken Chettinad with a side of paratha, and nothing […]

Published

on

This post is co-written by Chiranjeev Ghai, ML Engineer at zomato. zomato is a global food-tech company based in India.

Are you the kind of person who has very specific cravings? Maybe when the mood hits, you don’t want just any kind of Indian food—you want Chicken Chettinad with a side of paratha, and nothing else will hit the spot! To help picky eaters satisfy their cravings, we at zomato have recently added enhanced search engine capabilities to our restaurant aggregation and food delivery platform. These capabilities enable us to recommend restaurants to zomato users based on searches for specific dishes.

We power this functionality with machine learning (ML), using it to extract and structure text data from menu images. To develop this menu digitization technology, we partnered with Amazon ML Solutions Lab to explore the capabilities of the AWS ML Stack. This post summarizes how we used Amazon Textract and Amazon SageMaker to develop a customized menu digitization solution.

Extracting raw text from menus with Amazon Textract

The first component of this solution was to accurately extract all the text in the menu image. This process is known as optical character recognition (OCR). For our use case, we experimented with both in-house and commercial OCR solutions.

We first created an in-house OCR solution by stacking a pre-trained text detection model and a pre-trained text recognition model. The challenge with these models was that they were trained on a standard text dataset that didn’t match the eclectic fonts found in restaurant menus. To improve system performance, we fine-tuned these models by generating a dataset of 1.5 million synthetic text images that were more representative of text in menus.

After evaluating our in-house solution and several commercial OCR solutions, we found that Amazon Textract offers the best text recognition precision and recall. Restaurants often get creative when designing their menus, so OCR robustness was crucial for this use case. Amazon Textract particularly differentiated itself when processing menus with unique fonts, background images, and low image resolutions. Using it is as simple as making an API call:

#Python 3.6
import boto3
textract_client = boto3.client( 'textract', region_name = '' #insert the AWS region you're working in
)
textract_response = textract_client.detect_document_text( Document={ 'S3Object': { 'Bucket': '', #insert the name of the S3 bucket containing your image 'Name': '' #insert the S3 key of your image } }
) print(textract_response)

The following code is the Amazon Textract output for a sample image:

{'DocumentMetadata': {'Pages': 1}, 'Blocks': [{'BlockType': 'PAGE', 'Geometry': {'BoundingBox': {'Width': 1.0, 'Height': 1.0, 'Left': 0.0, 'Top': 0.0}, ... {'BlockType': 'WORD', 'Text': 'Dim', 'Geometry': {'BoundingBox': {'Width': 0.10242128372192383, 'Height': 0. 048968635499477386, 'Left': 0. 24052166938781738, 'Top': 0. 02556285448372364},
... 

The raw outputs are visualized by overlaying them on top of the image. The following image visualizes the preceding raw output. The black boxes are the text-detection bounding boxes provided by Amazon Textract. Extracted text is displayed on the right. Note the unconventional fonts, colors, and images on this menu.

The following image visualizes Amazon Textract outputs for a menu with a different design. Black boxes are the text-detection bounding boxes provided by Amazon Textract. Extracted text is displayed on the right. Again, this menu has unconventional fonts, colors, and images.

Using Amazon SageMaker to build a menu structure detector

The next component of this solution was to group the detections from Amazon Textract by menu section. This enabled our search engine to distinguish between entrees, desserts, beverages, and so on. We framed this as a computer vision problem—object detection, to be precise—and used Amazon SageMaker Ground Truth to collect training data. Ground Truth accelerated this process by providing a fully managed annotation tool that we customized to ask human annotators to draw bounding boxes around every menu section in the image. We used an annotation workforce from AWS Marketplace because this was a niche labeling task, and public labelers from Amazon Mechanical Turk didn’t perform well. With Ground Truth, it took just a few days and approximately $1,400 to label 4,086 images with triplicate redundancy.

With labeled data in hand, we faced a paradox of choice when selecting model-building approaches because object detection is such a thoroughly studied problem. Our choices included:

  • Removing low-confidence labels from the labeled dataset – Because even human annotators can make mistakes, Ground Truth calculates confidence scores for labels by having multiple annotators (for this use case, three) label the same image. Setting a higher confidence threshold for labels can decrease the noise in the training data at the expense of having less training data.
  • Data augmentation – Techniques for image data augmentation include horizontal flipping, cropping, shearing, and rotation. Data augmentation can make models more robust by increasing the amount of training data. However, excessive data augmentation may result in poor model convergence.
  • Feature engineering – From our experience in applying computer vision to processing menus, we had a variety of techniques in mind to emphasize or de-emphasize various aspects of the input images. For example, see the following images.

The following is the original image of a menu.

The following image shows the redacted image (overlay white boxes on a black background where text detections were found).

The following is a text cropped image. On a black background, the image has overlay crops from the original image where text detections were found.

The following is a single channel and text cropped image. The image is encoded as a single RGB channel (for this image, green). You can apply this with other transformations, in this case text cropping.

 

We also had the following additional model-building methods to choose from:

  • Model architectures like YOLO, SSD, and RCNN, with VGG or ResNet backbones – Each architecture has different trade-offs of model accuracy, inference time, model size, and more. For this use case, model accuracy was the most important metric because menu images were batch processed.
  • Using a model pre-trained on a general object detection task or starting from scratch – Transfer learning can be helpful when training complex models on small datasets. However, the task of detecting menu sections is very different from a general object detection task (for example, PASCAL VOC), so the pre-training may not be relevant.
  • Optimizer parameters – These include learning rate, momentum, regularization coefficients, and early stopping configuration.

With so many hyperparameters to consider, we turned to the automatic tuning feature of Amazon SageMaker to coordinate a massive tuning job across all these variables. The following code is an example of tuning a single model architecture and input data configuration:

import sagemaker
import boto3
from sagemaker.amazon.amazon_estimator import get_image_uri
from sagemaker.estimator import Estimator
from sagemaker.tuner import HyperparameterTuner, IntegerParameter, CategoricalParameter, ContinuousParameter
import itertools
from time import sleep #set to the region you're working in
REGION_NAME = ''
#set a S3 path for SageMaker to store the outputs of the training jobs S3_OUTPUT_PATH = ''
#set a S3 location for your training dataset, #assumed to be an augmented manifest file
#see: https://docs.aws.amazon.com/sagemaker/latest/dg/augmented-manifest.html
TRAIN_DATA_LOCATION = ''
#set a S3 location for your validation data, #assumed to be an augmented manifest file
VAL_DATA_LOCATION = ''
#specify which fields in the augmented manifest file are relevant for training
DATA_ATTRIBUTE_NAMES = [,]
#specify image shape
IMAGE_SHAPE = #specify label width
LABEL_WIDTH = #specify number of samples in the training dataset
NUM_TRAINING_SAMPLES = sgm_role = sagemaker.get_execution_role()
boto_session = boto3.session.Session( region_name = REGION_NAME
)
sgm_session = sagemaker.Session( boto_session = boto_session
)
training_image = get_image_uri( region_name = REGION_NAME, repo_name = 'object-detection', repo_version = 'latest'
) #set training job configuration
object_detection_estimator = Estimator( image_name = training_image, role = sgm_role, train_instance_count = 1, train_instance_type = 'ml.p3.2xlarge', train_volume_size = 50, train_max_run = 360000, input_mode = 'Pipe', output_path = S3_OUTPUT_PATH, sagemaker_session = sgm_session
) #set input data configuration
train_data = sagemaker.session.s3_input( s3_data = TRAIN_DATA_LOCATION, distribution = 'FullyReplicated', record_wrapping = 'RecordIO', s3_data_type = 'AugmentedManifestFile', attribute_names = DATA_ATTRIBUTE_NAMES
) val_data = sagemaker.session.s3_input( s3_data = VAL_DATA_LOCATION, distribution = 'FullyReplicated', record_wrapping = 'RecordIO', s3_data_type = 'AugmentedManifestFile', attribute_names = DATA_ATTRIBUTE_NAMES
) data_channels = { 'train': train_data, 'validation' : val_data
} #set static hyperparameters
#see: https://docs.aws.amazon.com/sagemaker/latest/dg/object-detection-api-config.html
static_hyperparameters = { 'num_classes' : 1, 'epochs' : 100, 'lr_scheduler_step' : '15,30', 'lr_scheduler_factor' : 0.1, 'overlap_threshold' : 0.5, 'nms_threshold' : 0.45, 'image_shape' : IMAGE_SHAPE, 'label_width' : LABEL_WIDTH, 'num_training_samples' : NUM_TRAINING_SAMPLES, 'early_stopping' : True, 'early_stopping_min_epochs' : 5, 'early_stopping_patience' : 1, 'early_stopping_tolerance' : 0.05,
} #set ranges for tunable hyperparameters
hyperparameter_ranges = { 'learning_rate': ContinuousParameter( min_value = 1e-5, max_value = 1e-2, scaling_type = 'Auto' ), 'mini_batch_size': IntegerParameter( min_value = 8, max_value = 64, scaling_type = 'Auto' )
} #Not all hyperparameters are feasible to tune directly
#see: https://docs.aws.amazon.com/sagemaker/latest/dg/object-detection-tuning.html
#For these we run model tuning jobs in parallel using a for loop
#We take this approach for tuning over different model architectures #and different feature engineering configurations
use_pretrained_options = [0, 1]
base_network_options = ['resnet-50', 'vgg-16'] for use_pretrained, base_network in itertools.product(use_pretrained_options, base_network_options): static_hyperparameter_configuration = { **static_hyperparameters, 'use_pretrained_model' : use_pretrained, 'base_network' : base_network } object_detection_estimator.set_hyperparameters( **static_hyperparameter_configuration ) tuner = HyperparameterTuner( estimator = object_detection_estimator, objective_metric_name = 'validation:mAP', strategy = 'Bayesian', hyperparameter_ranges = hyperparameter_ranges, max_jobs = 24, max_parallel_jobs = 2, early_stopping_type = 'Auto', ) tuner.fit( inputs = data_channels ) print(f'Started tuning job: {tuner.latest_tuning_job.name}') #wait a bit before starting next job so auto generated names don't conflict sleep(60)

This code uses version 1.72.0 of the Amazon SageMaker Python SDK, which is the default version installed in Amazon SageMaker notebook instances. Version 2.X introduces breaking changes. For more information, see Use Version 2.x of the SageMaker Python SDK.

We used powerful GPU hardware (p3.2xlarge instances), and it took us just 1 week and approximately $1,500 to explore 455 unique parameter configurations. Of these configurations, Amazon SageMaker found that a fine-tuned Faster R-CNN model with text cropping performed the best, with a mean average precision score of 0.93. This aligned with results from our prior work in this space, which found that two-stage detectors generally outperform single-stage detectors in processing menus.

The following is an example of how the object detection model processed a menu. In this image, the purple boxes are the predicted bounding boxes from the menu section detection model. Black boxes are the text detection bounding boxes provided by Amazon Textract.

Using Amazon SageMaker to build rule- and ML-based text classifiers

The final component in the solution was a layer of text classification. To enable our enhanced search functionality, we had to know if each detection within a menu section was the menu section title, name of a dish, price of a dish, or something else (such as a description of a dish or the name of the restaurant). To this end, we developed a hybrid rule- and ML-based text classification system.

The first step of the classification was to use a rule to determine if a detection was a price or not. This rule simply calculated the proportion of numeric characters in the detection. If the proportion was greater than 40%, the detection was classified as a price. Although simple, this classifier worked well in practice. We used Amazon SageMaker notebook instances as a convenient interactive environment to develop this and other rules.

After the prices were filtered out, the remaining detections were classified as dish or not dish. From our experience in processing menus, we intuitively knew that in many cases, the location of prices was sufficient to do this classification. For these menus, dishes and prices are listed side by side, so simply classifying detections located to the left of prices as dishes worked well.

The following example shows how the rules-based text classification system processed a menu. Green boxes are detections classified as dishes (by the price location rule). Red boxes are detections classified as not dishes (by the price location rule). Blue boxes are detections classified as prices. Final dish detections are on the right.

Some menus might include lengthy dish descriptions or may not list prices next to individual dishes. These menus violate the assumptions of the price location rules, so we turned to model-based text classification. We used Amazon SageMaker training jobs to experiment with many modeling approaches in parallel, including an XGBoost model trained on hashed word count vectors. In the end, we found that a fine-tuned BERT model from GluonNLP achieved the best performance with an AUROC score of 0.86.

The following image is an example of how the model-based text classification system processed a menu. Green boxes are detections classified as dishes (by the BERT model). Red boxes are detections classified as not dishes (by the BERT model). Blue boxes are detections classified as prices. The final dish detections are on the right.

Of the remaining detections (those not classified as prices or dishes), a final round of classification identified menu section titles. We created features that captured the font size of the detection, the location of the detection on the menu, and the length of the words within the detection. We used these features as inputs to a logistic regression model that predicted if a detection is a menu section title or not.

Key features of Amazon SageMaker

In the end, we found that doing OCR was as simple as making an API call to Amazon Textract. However, our use case required additional customization. We selected Amazon SageMaker as an ML platform to develop this customization because it offered several key features:

  • Amazon SageMaker Notebooks made it easy to spin up Jupyter notebook environments for prototyping and testing rules and models.
  • Ground Truth helped us build and deploy a custom image annotation tool with no front-end experience required.
  • Amazon SageMaker automatic tuning enabled us to run massive hyperparameter tuning jobs on powerful hardware, and included an intuitive interface for tracking the results of hundreds of experiments. You can implement tuning jobs with early stopping conditions, which makes experimentation cost-effective.

Amazon SageMaker offers additional integration benefits from including all the preceding features in a single platform:

  • Amazon SageMaker Notebooks come pre-installed with all the dependencies needed to build models that can be optimized with automatic tuning.
  • Ground Truth offers easy access to labelers from Mechanical Turk or AWS Marketplace.
  • Automatic tuning can directly ingest the manifest files created by Amazon SageMaker Ground Truth.

Putting it all together

Our menu digitization system can extract text from images of menus, group it by menu section, extract the title of the section, extract the dishes within each section, and pair each dish with its price. The following is a visualization of the end-to-end solution.

The workflow contains the following steps:

  1. The input is an image of a menu.
  2. Amazon Textract performs OCR on the input image.
  3. An ML-based computer vision model predicts bounding boxes for menu sections in the menu image.
  4. A rules-based classifier classifies Amazon Textract detections as price or not price.
  5. A rules-based classifier (5a) attempts to use the location of price detections to classify the not price detections as dish or not dish. If this rule doesn’t successfully classify most of the detections on the page, an ML-based classifier is used instead (5b).
  6. The ML-based classifier uses hand-crafted features to classify not dish detections as menu section title or not menu section title.
  7.  The menu text is structured by combining the menu section detections and the text classification results.

The following image visualizes a sample output of the system. Green boxes are detections classified as dishes. Blue boxes are detections classified as prices. Yellow boxes are detections classified as menu section titles. Purple boxes are predicted menu section bounding boxes.

The following code is the structured output:

[ { "title":{ "text":"Shrimp Dishes" }, "dishes":[ { "text":"Shrimp Masala", "price":{ "text":"140" } }, { "text":"Shrimp Biryani", "price":{ "text":"170" } }, { "text":"Shrimp Pulav", "price":{ "text":"160" } } ] }, ...
]

Conclusion

We built a system that uses ML to digitize menus without any human input required. This system will improve user experience by powering new features such as advanced dish search and review highlight verification. Our content team will also use it to accelerate creating menus for online ordering.

To explore these capabilities of Amazon Textract and Amazon SageMaker in more depth, see Automatically extract text and structured data from documents with Amazon Textract and Amazon SageMaker Automatic Model Tuning: Using Machine Learning for Machine Learning.

The Amazon ML Solutions Lab helped us accelerate our use of ML by pairing our team with ML experts. The ML Solutions Lab brings to every customer engagement learnings from more than 20 years of Amazon’s ML innovations in areas such as fulfillment and logistics, personalization and recommendations, computer vision and translation, fraud prevention, forecasting, and supply chain optimization. To learn more about the AWS ML Solutions Lab, contact your account manager or visit Amazon Machine Learning Solutions Lab.


About the Authors

Chiranjeev Ghai is a Machine Learning Engineer. In his current role, he has been aiding automation at zomato by leveraging a wide variety of ML optimisations ranging from Image Classification, Product Recommendation, and Text Detection. When not building models, he likes to spend his time playing video games at home.

Ryan Cheng is a Deep Learning Architect in the Amazon ML Solutions Lab. He has worked on a wide range of ML use cases from sports analytics to optical character recognition. In his spare time, Ryan enjoys cooking.

Andrew Ang is a Deep Learning Architect at the Amazon ML Solutions Lab, where he helps AWS customers identify and build AI/ML solutions to address their business problems.

Vinayak Arannil is a Data Scientist at the Amazon Machine Learning Solutions Lab. He has worked on various domains of data science like computer vision, natural language processing, recommendation systems, etc.

Source: https://aws.amazon.com/blogs/machine-learning/zomato-digitizes-menus-using-amazon-textract-and-amazon-sagemaker/

Continue Reading

AI

zomato digitizes menus using Amazon Textract and Amazon SageMaker

This post is co-written by Chiranjeev Ghai, ML Engineer at zomato. zomato is a global food-tech company based in India. Are you the kind of person who has very specific cravings? Maybe when the mood hits, you don’t want just any kind of Indian food—you want Chicken Chettinad with a side of paratha, and nothing […]

Published

on

This post is co-written by Chiranjeev Ghai, ML Engineer at zomato. zomato is a global food-tech company based in India.

Are you the kind of person who has very specific cravings? Maybe when the mood hits, you don’t want just any kind of Indian food—you want Chicken Chettinad with a side of paratha, and nothing else will hit the spot! To help picky eaters satisfy their cravings, we at zomato have recently added enhanced search engine capabilities to our restaurant aggregation and food delivery platform. These capabilities enable us to recommend restaurants to zomato users based on searches for specific dishes.

We power this functionality with machine learning (ML), using it to extract and structure text data from menu images. To develop this menu digitization technology, we partnered with Amazon ML Solutions Lab to explore the capabilities of the AWS ML Stack. This post summarizes how we used Amazon Textract and Amazon SageMaker to develop a customized menu digitization solution.

Extracting raw text from menus with Amazon Textract

The first component of this solution was to accurately extract all the text in the menu image. This process is known as optical character recognition (OCR). For our use case, we experimented with both in-house and commercial OCR solutions.

We first created an in-house OCR solution by stacking a pre-trained text detection model and a pre-trained text recognition model. The challenge with these models was that they were trained on a standard text dataset that didn’t match the eclectic fonts found in restaurant menus. To improve system performance, we fine-tuned these models by generating a dataset of 1.5 million synthetic text images that were more representative of text in menus.

After evaluating our in-house solution and several commercial OCR solutions, we found that Amazon Textract offers the best text recognition precision and recall. Restaurants often get creative when designing their menus, so OCR robustness was crucial for this use case. Amazon Textract particularly differentiated itself when processing menus with unique fonts, background images, and low image resolutions. Using it is as simple as making an API call:

#Python 3.6
import boto3
textract_client = boto3.client( 'textract', region_name = '' #insert the AWS region you're working in
)
textract_response = textract_client.detect_document_text( Document={ 'S3Object': { 'Bucket': '', #insert the name of the S3 bucket containing your image 'Name': '' #insert the S3 key of your image } }
) print(textract_response)

The following code is the Amazon Textract output for a sample image:

{'DocumentMetadata': {'Pages': 1}, 'Blocks': [{'BlockType': 'PAGE', 'Geometry': {'BoundingBox': {'Width': 1.0, 'Height': 1.0, 'Left': 0.0, 'Top': 0.0}, ... {'BlockType': 'WORD', 'Text': 'Dim', 'Geometry': {'BoundingBox': {'Width': 0.10242128372192383, 'Height': 0. 048968635499477386, 'Left': 0. 24052166938781738, 'Top': 0. 02556285448372364},
... 

The raw outputs are visualized by overlaying them on top of the image. The following image visualizes the preceding raw output. The black boxes are the text-detection bounding boxes provided by Amazon Textract. Extracted text is displayed on the right. Note the unconventional fonts, colors, and images on this menu.

The following image visualizes Amazon Textract outputs for a menu with a different design. Black boxes are the text-detection bounding boxes provided by Amazon Textract. Extracted text is displayed on the right. Again, this menu has unconventional fonts, colors, and images.

Using Amazon SageMaker to build a menu structure detector

The next component of this solution was to group the detections from Amazon Textract by menu section. This enabled our search engine to distinguish between entrees, desserts, beverages, and so on. We framed this as a computer vision problem—object detection, to be precise—and used Amazon SageMaker Ground Truth to collect training data. Ground Truth accelerated this process by providing a fully managed annotation tool that we customized to ask human annotators to draw bounding boxes around every menu section in the image. We used an annotation workforce from AWS Marketplace because this was a niche labeling task, and public labelers from Amazon Mechanical Turk didn’t perform well. With Ground Truth, it took just a few days and approximately $1,400 to label 4,086 images with triplicate redundancy.

With labeled data in hand, we faced a paradox of choice when selecting model-building approaches because object detection is such a thoroughly studied problem. Our choices included:

  • Removing low-confidence labels from the labeled dataset – Because even human annotators can make mistakes, Ground Truth calculates confidence scores for labels by having multiple annotators (for this use case, three) label the same image. Setting a higher confidence threshold for labels can decrease the noise in the training data at the expense of having less training data.
  • Data augmentation – Techniques for image data augmentation include horizontal flipping, cropping, shearing, and rotation. Data augmentation can make models more robust by increasing the amount of training data. However, excessive data augmentation may result in poor model convergence.
  • Feature engineering – From our experience in applying computer vision to processing menus, we had a variety of techniques in mind to emphasize or de-emphasize various aspects of the input images. For example, see the following images.

The following is the original image of a menu.

The following image shows the redacted image (overlay white boxes on a black background where text detections were found).

The following is a text cropped image. On a black background, the image has overlay crops from the original image where text detections were found.

The following is a single channel and text cropped image. The image is encoded as a single RGB channel (for this image, green). You can apply this with other transformations, in this case text cropping.

 

We also had the following additional model-building methods to choose from:

  • Model architectures like YOLO, SSD, and RCNN, with VGG or ResNet backbones – Each architecture has different trade-offs of model accuracy, inference time, model size, and more. For this use case, model accuracy was the most important metric because menu images were batch processed.
  • Using a model pre-trained on a general object detection task or starting from scratch – Transfer learning can be helpful when training complex models on small datasets. However, the task of detecting menu sections is very different from a general object detection task (for example, PASCAL VOC), so the pre-training may not be relevant.
  • Optimizer parameters – These include learning rate, momentum, regularization coefficients, and early stopping configuration.

With so many hyperparameters to consider, we turned to the automatic tuning feature of Amazon SageMaker to coordinate a massive tuning job across all these variables. The following code is an example of tuning a single model architecture and input data configuration:

import sagemaker
import boto3
from sagemaker.amazon.amazon_estimator import get_image_uri
from sagemaker.estimator import Estimator
from sagemaker.tuner import HyperparameterTuner, IntegerParameter, CategoricalParameter, ContinuousParameter
import itertools
from time import sleep #set to the region you're working in
REGION_NAME = ''
#set a S3 path for SageMaker to store the outputs of the training jobs S3_OUTPUT_PATH = ''
#set a S3 location for your training dataset, #assumed to be an augmented manifest file
#see: https://docs.aws.amazon.com/sagemaker/latest/dg/augmented-manifest.html
TRAIN_DATA_LOCATION = ''
#set a S3 location for your validation data, #assumed to be an augmented manifest file
VAL_DATA_LOCATION = ''
#specify which fields in the augmented manifest file are relevant for training
DATA_ATTRIBUTE_NAMES = [,]
#specify image shape
IMAGE_SHAPE = #specify label width
LABEL_WIDTH = #specify number of samples in the training dataset
NUM_TRAINING_SAMPLES = sgm_role = sagemaker.get_execution_role()
boto_session = boto3.session.Session( region_name = REGION_NAME
)
sgm_session = sagemaker.Session( boto_session = boto_session
)
training_image = get_image_uri( region_name = REGION_NAME, repo_name = 'object-detection', repo_version = 'latest'
) #set training job configuration
object_detection_estimator = Estimator( image_name = training_image, role = sgm_role, train_instance_count = 1, train_instance_type = 'ml.p3.2xlarge', train_volume_size = 50, train_max_run = 360000, input_mode = 'Pipe', output_path = S3_OUTPUT_PATH, sagemaker_session = sgm_session
) #set input data configuration
train_data = sagemaker.session.s3_input( s3_data = TRAIN_DATA_LOCATION, distribution = 'FullyReplicated', record_wrapping = 'RecordIO', s3_data_type = 'AugmentedManifestFile', attribute_names = DATA_ATTRIBUTE_NAMES
) val_data = sagemaker.session.s3_input( s3_data = VAL_DATA_LOCATION, distribution = 'FullyReplicated', record_wrapping = 'RecordIO', s3_data_type = 'AugmentedManifestFile', attribute_names = DATA_ATTRIBUTE_NAMES
) data_channels = { 'train': train_data, 'validation' : val_data
} #set static hyperparameters
#see: https://docs.aws.amazon.com/sagemaker/latest/dg/object-detection-api-config.html
static_hyperparameters = { 'num_classes' : 1, 'epochs' : 100, 'lr_scheduler_step' : '15,30', 'lr_scheduler_factor' : 0.1, 'overlap_threshold' : 0.5, 'nms_threshold' : 0.45, 'image_shape' : IMAGE_SHAPE, 'label_width' : LABEL_WIDTH, 'num_training_samples' : NUM_TRAINING_SAMPLES, 'early_stopping' : True, 'early_stopping_min_epochs' : 5, 'early_stopping_patience' : 1, 'early_stopping_tolerance' : 0.05,
} #set ranges for tunable hyperparameters
hyperparameter_ranges = { 'learning_rate': ContinuousParameter( min_value = 1e-5, max_value = 1e-2, scaling_type = 'Auto' ), 'mini_batch_size': IntegerParameter( min_value = 8, max_value = 64, scaling_type = 'Auto' )
} #Not all hyperparameters are feasible to tune directly
#see: https://docs.aws.amazon.com/sagemaker/latest/dg/object-detection-tuning.html
#For these we run model tuning jobs in parallel using a for loop
#We take this approach for tuning over different model architectures #and different feature engineering configurations
use_pretrained_options = [0, 1]
base_network_options = ['resnet-50', 'vgg-16'] for use_pretrained, base_network in itertools.product(use_pretrained_options, base_network_options): static_hyperparameter_configuration = { **static_hyperparameters, 'use_pretrained_model' : use_pretrained, 'base_network' : base_network } object_detection_estimator.set_hyperparameters( **static_hyperparameter_configuration ) tuner = HyperparameterTuner( estimator = object_detection_estimator, objective_metric_name = 'validation:mAP', strategy = 'Bayesian', hyperparameter_ranges = hyperparameter_ranges, max_jobs = 24, max_parallel_jobs = 2, early_stopping_type = 'Auto', ) tuner.fit( inputs = data_channels ) print(f'Started tuning job: {tuner.latest_tuning_job.name}') #wait a bit before starting next job so auto generated names don't conflict sleep(60)

This code uses version 1.72.0 of the Amazon SageMaker Python SDK, which is the default version installed in Amazon SageMaker notebook instances. Version 2.X introduces breaking changes. For more information, see Use Version 2.x of the SageMaker Python SDK.

We used powerful GPU hardware (p3.2xlarge instances), and it took us just 1 week and approximately $1,500 to explore 455 unique parameter configurations. Of these configurations, Amazon SageMaker found that a fine-tuned Faster R-CNN model with text cropping performed the best, with a mean average precision score of 0.93. This aligned with results from our prior work in this space, which found that two-stage detectors generally outperform single-stage detectors in processing menus.

The following is an example of how the object detection model processed a menu. In this image, the purple boxes are the predicted bounding boxes from the menu section detection model. Black boxes are the text detection bounding boxes provided by Amazon Textract.

Using Amazon SageMaker to build rule- and ML-based text classifiers

The final component in the solution was a layer of text classification. To enable our enhanced search functionality, we had to know if each detection within a menu section was the menu section title, name of a dish, price of a dish, or something else (such as a description of a dish or the name of the restaurant). To this end, we developed a hybrid rule- and ML-based text classification system.

The first step of the classification was to use a rule to determine if a detection was a price or not. This rule simply calculated the proportion of numeric characters in the detection. If the proportion was greater than 40%, the detection was classified as a price. Although simple, this classifier worked well in practice. We used Amazon SageMaker notebook instances as a convenient interactive environment to develop this and other rules.

After the prices were filtered out, the remaining detections were classified as dish or not dish. From our experience in processing menus, we intuitively knew that in many cases, the location of prices was sufficient to do this classification. For these menus, dishes and prices are listed side by side, so simply classifying detections located to the left of prices as dishes worked well.

The following example shows how the rules-based text classification system processed a menu. Green boxes are detections classified as dishes (by the price location rule). Red boxes are detections classified as not dishes (by the price location rule). Blue boxes are detections classified as prices. Final dish detections are on the right.

Some menus might include lengthy dish descriptions or may not list prices next to individual dishes. These menus violate the assumptions of the price location rules, so we turned to model-based text classification. We used Amazon SageMaker training jobs to experiment with many modeling approaches in parallel, including an XGBoost model trained on hashed word count vectors. In the end, we found that a fine-tuned BERT model from GluonNLP achieved the best performance with an AUROC score of 0.86.

The following image is an example of how the model-based text classification system processed a menu. Green boxes are detections classified as dishes (by the BERT model). Red boxes are detections classified as not dishes (by the BERT model). Blue boxes are detections classified as prices. The final dish detections are on the right.

Of the remaining detections (those not classified as prices or dishes), a final round of classification identified menu section titles. We created features that captured the font size of the detection, the location of the detection on the menu, and the length of the words within the detection. We used these features as inputs to a logistic regression model that predicted if a detection is a menu section title or not.

Key features of Amazon SageMaker

In the end, we found that doing OCR was as simple as making an API call to Amazon Textract. However, our use case required additional customization. We selected Amazon SageMaker as an ML platform to develop this customization because it offered several key features:

  • Amazon SageMaker Notebooks made it easy to spin up Jupyter notebook environments for prototyping and testing rules and models.
  • Ground Truth helped us build and deploy a custom image annotation tool with no front-end experience required.
  • Amazon SageMaker automatic tuning enabled us to run massive hyperparameter tuning jobs on powerful hardware, and included an intuitive interface for tracking the results of hundreds of experiments. You can implement tuning jobs with early stopping conditions, which makes experimentation cost-effective.

Amazon SageMaker offers additional integration benefits from including all the preceding features in a single platform:

  • Amazon SageMaker Notebooks come pre-installed with all the dependencies needed to build models that can be optimized with automatic tuning.
  • Ground Truth offers easy access to labelers from Mechanical Turk or AWS Marketplace.
  • Automatic tuning can directly ingest the manifest files created by Amazon SageMaker Ground Truth.

Putting it all together

Our menu digitization system can extract text from images of menus, group it by menu section, extract the title of the section, extract the dishes within each section, and pair each dish with its price. The following is a visualization of the end-to-end solution.

The workflow contains the following steps:

  1. The input is an image of a menu.
  2. Amazon Textract performs OCR on the input image.
  3. An ML-based computer vision model predicts bounding boxes for menu sections in the menu image.
  4. A rules-based classifier classifies Amazon Textract detections as price or not price.
  5. A rules-based classifier (5a) attempts to use the location of price detections to classify the not price detections as dish or not dish. If this rule doesn’t successfully classify most of the detections on the page, an ML-based classifier is used instead (5b).
  6. The ML-based classifier uses hand-crafted features to classify not dish detections as menu section title or not menu section title.
  7.  The menu text is structured by combining the menu section detections and the text classification results.

The following image visualizes a sample output of the system. Green boxes are detections classified as dishes. Blue boxes are detections classified as prices. Yellow boxes are detections classified as menu section titles. Purple boxes are predicted menu section bounding boxes.

The following code is the structured output:

[ { "title":{ "text":"Shrimp Dishes" }, "dishes":[ { "text":"Shrimp Masala", "price":{ "text":"140" } }, { "text":"Shrimp Biryani", "price":{ "text":"170" } }, { "text":"Shrimp Pulav", "price":{ "text":"160" } } ] }, ...
]

Conclusion

We built a system that uses ML to digitize menus without any human input required. This system will improve user experience by powering new features such as advanced dish search and review highlight verification. Our content team will also use it to accelerate creating menus for online ordering.

To explore these capabilities of Amazon Textract and Amazon SageMaker in more depth, see Automatically extract text and structured data from documents with Amazon Textract and Amazon SageMaker Automatic Model Tuning: Using Machine Learning for Machine Learning.

The Amazon ML Solutions Lab helped us accelerate our use of ML by pairing our team with ML experts. The ML Solutions Lab brings to every customer engagement learnings from more than 20 years of Amazon’s ML innovations in areas such as fulfillment and logistics, personalization and recommendations, computer vision and translation, fraud prevention, forecasting, and supply chain optimization. To learn more about the AWS ML Solutions Lab, contact your account manager or visit Amazon Machine Learning Solutions Lab.


About the Authors

Chiranjeev Ghai is a Machine Learning Engineer. In his current role, he has been aiding automation at zomato by leveraging a wide variety of ML optimisations ranging from Image Classification, Product Recommendation, and Text Detection. When not building models, he likes to spend his time playing video games at home.

Ryan Cheng is a Deep Learning Architect in the Amazon ML Solutions Lab. He has worked on a wide range of ML use cases from sports analytics to optical character recognition. In his spare time, Ryan enjoys cooking.

Andrew Ang is a Deep Learning Architect at the Amazon ML Solutions Lab, where he helps AWS customers identify and build AI/ML solutions to address their business problems.

Vinayak Arannil is a Data Scientist at the Amazon Machine Learning Solutions Lab. He has worked on various domains of data science like computer vision, natural language processing, recommendation systems, etc.

Source: https://aws.amazon.com/blogs/machine-learning/zomato-digitizes-menus-using-amazon-textract-and-amazon-sagemaker/

Continue Reading
AI2 hours ago

5 Work From Home Office Essentials

AI2 hours ago

zomato digitizes menus using Amazon Textract and Amazon SageMaker

AI2 hours ago

zomato digitizes menus using Amazon Textract and Amazon SageMaker

AI2 hours ago

zomato digitizes menus using Amazon Textract and Amazon SageMaker

AI2 hours ago

How to Improve Your Supply Chain With Deep Reinforcement Learning

AI2 hours ago

Video streaming and deep learning: Using Amazon Kinesis Video Streams with Deep Java Library

AI2 hours ago

Video streaming and deep learning: Using Amazon Kinesis Video Streams with Deep Java Library

AI2 hours ago

Video streaming and deep learning: Using Amazon Kinesis Video Streams with Deep Java Library

AI2 hours ago

Video streaming and deep learning: Using Amazon Kinesis Video Streams with Deep Java Library

AI5 hours ago

Conversation Designers: who are they and what do they do?

AI5 hours ago

Automating Bot Testing at Haptik

AI7 hours ago

Why Facebook’s New Machine Translation Model is a Great Step for AI

AI1 day ago

Bringing real-time machine learning-powered insights to rugby using Amazon SageMaker

AI1 day ago

Bringing real-time machine learning-powered insights to rugby using Amazon SageMaker

AI1 day ago

Bringing real-time machine learning-powered insights to rugby using Amazon SageMaker

AI1 day ago

Bringing real-time machine learning-powered insights to rugby using Amazon SageMaker

AI1 day ago

Bringing real-time machine learning-powered insights to rugby using Amazon SageMaker

AI1 day ago

Bringing real-time machine learning-powered insights to rugby using Amazon SageMaker

AI1 day ago

Bringing real-time machine learning-powered insights to rugby using Amazon SageMaker

AI1 day ago

Bringing real-time machine learning-powered insights to rugby using Amazon SageMaker

Trending