ICML 2019 Expo Demo

June 9, 2019

Expo Schedule »

Imagine-api, a python3 SDK for performing predictions using doc.ai tensorflow models

Sponsor: doc.ai

Organizers:
Wan Zhu (doc.ai)

Presenters:
Brian Hsu (doc.ai)

Abstract:

We have developed three image-based AI models, namely phenomenal-face, happy-face and manna. Specifically, phenomenal-face takes selfie images as input and predicts age, gender, height and BMI of the user; happy-face also takes selfie images as input and predicts the user’s mood, i.e. happy, calm, upset, surprised or sad; and manna takes food images as input and predicts whether it is either healthy or unhealthy. In order to enable users access to these models, we developed a python client SDK named Imagine-api. Users can send input data, an image, using the Imagine-api directly to our tensorflow models and receive predictions as a JSON object. The advantage of Imagine-api is to provide an open source tool for users to easily access our health-related AI models.