This repository contains code and documentation for training a ResNet model on a human action recognition dataset. The goal is to accurately classify different human actions such as eating, sitting, calling, and more. The experiment includes hyperparameter search, train-validation-test split, and the use of transfer learning.
Show your appreciation: If you find this project useful, please consider showing your support by starring it on GitHub. Your support means a lot! ⭐
View experiments on W&B
Check src/Data/README.md file for instruction on how to setup dataset
export WANDB_API_KEY=<your_key>
echo $WANDB_API_KEY
python train.py --n_epochs 10 --batch_size 128 --learning_rate 0.001
View train.py file for more detals
Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0)