Repository used to Train and Test Neural Nets for Autism Spectrum Disorder (detects Therapist, Toddler and Teddy Bear)

⚠️ This repository is incomplete, some expected features are not here, like validation.

⚠️ This repository served as an isolated place to train and test NNs that are being used in the web app repository. For now, it uses YOLOv5 to detect interactions between actors in a ASD therapy session.

Install

1. Clone YOLOv5 and annotate its path.
2. Clone this repo and install its requirements.
   1. Update YOLOv5 path variable in the config file.

# Clone YOLOv5 and store its home location since it is not pip installable
git clone https://github.com/ultralytics/yolov5.git

# Clone and install this repo
git clone https://github.com/ASDDataMining/therapy-aid-nn.git
cd therapy-aid-nn
# python3 -m venv venv; source venv/bin/activate  # optional: python virtual environment
pip install -r requirements.txt
# Then go to cfg/yolo_location.cfg and change yolov5's path

Usage

• After installation you can adjust the configuration files to your needs and run any of the corresponding .py files: train.py and detect.py.
• If wandb was installed correctly you can make logs in it
• For downloading labelbox images with the script provided you must create a .env file at the repo top-level and edit LABELBOX_API=<your-labelbox-api>.