Object Tracker

Segmentation module and tracking.

---

Requirements

This project has the following requirements:

1. Python: Version 3.10
2. CUDA: Version 12.1 (for GPU support)
3. Dependencies:
   • PyTorch 2.2.0 and related libraries
   • Torch Geometric and extensions
   • Git-based libraries:
     • GroundingDINO
     • Segment Anything Model (SAM)
     • XMem
   • Core libraries like numpy, opencv-python-headless, scipy, scikit-learn, and open3d

---

Installation

Follow these steps to install and set up the project:

1. Clone the Repository

Clone the project repository to your local machine:

git clone <repository-url>
cd <repository-directory>

2. Set Up a Python Environment

1. Ensure you have Conda installed on your system. If not, you can download it from Miniconda or Anaconda.
2. Create a new Conda environment with Python 3.10:

conda create -n tracker python=3.10 -y
conda activate tracker

3. Install the project dependencies

Setup the environment by running:

bash setup_env.sh

Install XMem:

git clone https://github.com/hkchengrex/XMem.git
cd XMem
pip install -r requirements.txt
cd ..

Remember to export XMem path to your PYTHONPATH variable:

export PYTHONPATH=$PYTHONPATH:<path-to-folder>/object_tracker
export PYTHONPATH=$PYTHONPATH:<path-to-folder>/object_tracker/XMem

4. Verify PyTorch, CUDA and Additional Dependencies

Make sure the environemtn Test that PyTorch and CUDA are installed correctly:

python -c "import torch
print(f'Torch Version: {torch.__version__}')
print(f'CUDA Available: {torch.cuda.is_available()}')
"

Expected output:

Torch Version: 2.2.0+cu121
CUDA Available: True

Test that GroundingDINO is installed correctly:

python -c "
from groundingdino.util.inference import load_model
print('GroundingDINO installed successfully.')
"

Test that SAM is installed correctly:

python -c "
from segment_anything import SamPredictor
print('SAM installed successfully.')
"

Test that XMem is installed correctly:

python -c "
from XMem.model.network import XMem
print('XMem installed successfully.')
"