We developed a mobile facial recognition app for an eyewear brand that builds custom frames. We needed to create a system to estimate the inter-pupillary distance of a person based on mobile images.
We used AWS Rekognition to detect the face pose, landmarks and the bounding box on each picture. We then inferred the interpupillary distance with a computer vision algorithm, cross-validated results with a dataset, and iterated the process to improve accuracy.
We worked with a Tax firm to classify credit card transactions to automate and improve the quality of tax form deliverables. We first completed preprocessing to mitigate entropy in transaction descriptions. We then used AWS Comprehend to feature enhance the transaction descriptions.
After completing the model we deployed it on AWS Sagemaker which allows us to build, train, and deploy new models as more data becomes available.
Image recognition for a global athletic brand
We helped one of the world’s largest athletic retailers deliver the hottest trends by influencing product decisions earlier in the production cycle. Our team built an algorithm to look through a library of thousands of images to identify stylistic trends and colors.
We combined picture metadata and collected garment characteristics by using the detection features in AWS Rekognition with APIs and custom models running on top of AWS SageMaker.
Talk to us about building AI and ML models using AWS tools
We will get in touch with you shortly.