This project contains implementation of Human Pose Estimation with a custom convolutional block and a pre-trained architecture (ResNet)

The dataset used is COCO2017, downloaded from kaggle - https://www.kaggle.com/datasets/awsaf49/coco-2017-dataset

The project structure is organized in two formats -

1. .py files which includes data.py (implementation of the dataset), metrics.py (implementation of the loss function and PCK accuracy metric), conv.py (containing our convolutional block), utils.py (helper functions to initialize the hyperparameters) and train.py (to train the final model)

2. .ipynb file, a jupyter notebook which contains dataset preprocessing, convolutional model creation to training and inference of the code and can be run at once

The code is fairly self-explanatory, if you encounter any errors or have any queries, reach out at [email protected]