Tristar: Support Code

This repository contains the support code for "A Trimodal Dataset: RGB, Thermal, and Depth for Human Segmentation and Action Recognition". It contains data loaders as well as code for training and testing the described benchmark models.

Installation

Add here instructions on how to clone this repository and install any dependencies. This might be as simple as:

git clone https://github.com/anonymous/tristar.git
cd tristar
pip install -r requirements.txt

Usage

Download the dataset from zenodo, unpack and move it to data:

mkdir data
cd data
wget https://zenodo.org/record/7996570/files/tristar.zip
unzip tristar.zip
cd ..

Create your environment and install the requirements, for example:

conda create -n tristar python=3.8
conda activate tristar
pip install -r requirements.txt

To run every valid model modality combination with 10 epochs use:

chmod +x train_all.sh
./train_all.sh

Finally, to create json files with the test scores:

python -m tristar.test_best