Human Facial Emotion Recognition (HFER): Back End

Introduction

HFER is a Python project that leverages convolutional neural networks to detect emotions from human faces which are extracted from images uploaded by a user. It can recognize the following seven emotions: surprise, fear, disgust, happiness, anger, sadness, and neutral. We used multiclass image classification, a type of supervised deep learning algorithm. This is the code repo for the back-end implementation. For general information about the project, see the project's main README.

Usage

REST API

HFER's REST API was written using FastAPI and the container has been deployed on Google Cloud Run which can be accessed here.

Note: Cold starts are enabled by default on GCR, so the container may take up to a few minutes to start the first time it is accessed.

How to use the API

❗️❗️❗️ TO DO ❗️❗️❗️

Step-by-step guide to using the API and a link to the API documentation once generated

Web UI

HFER has a web interface written using Streamlit. The code resides in its own separate repo here. It is currently deployed here.

Installation (For Development):

Install Locally

1. Clone the repo.
2. Download the trained model file here and copy it into the models directory.
3. Install the hfer package in editable mode with:

   pip install -e .

4. From the hfer/hfer directory run:

   uvicorn hfer.server.main_server:app --reload

Install Using Docker

1. Follow steps 1-2 above.
2. Build the docker image with:

   docker build -t hfer-api .

3. Run the container with:

   docker run -p 8000:8000 hfer-api

Note: If you would like to deploy the front-end locally as well, follow the instructions in the hfer_front repo's README.

Contributors

HFER was developed by two friends, Farid and Nathan, with significant input and guidance from a third friend: Or.