Deformable Capsule Network for Object Detection

This project implements a Deformable Capsule Network (DeformCaps) for object detection using PyTorch, as described in the paper "Deformable Capsules for Object Detection".

Overview

Deformable Capsules (DeformCaps) are designed to address the challenge of object detection in computer vision. This implementation includes:

A novel capsule structure (SplitCaps)
A dynamic routing algorithm (SE-Routing)
Training and evaluation on the MS COCO dataset

Installation

Clone the repository:

git clone https://github.com/naivoder/DeformableCapsuleNetwork.git
cd DeformableCapsuleNetwork

Install the required dependencies:
```
pip install -r requirements.txt
```

Usage

Download and extract the MS COCO dataset:
- This will be handled automatically by the script.
Run the training and evaluation script:
```
python main.py
```

This will download the MS COCO dataset, train the DeforCaps Network, and evaluate its performance.

Note

Make sure you have sufficient disk space and memory for downloading and processing the MS COCO dataset (~20 Gb).