- TODO: Introduction to project
The only prerequisite is git in order to clone the repository. The project was tested using a Python 3.8.5 interpreter.
- Open a terminal window
- Clone repository using
git clone https://github.com/AndreasKaratzas/faster-rcnn.git
- Navigate to project directory with
cd faster-rcnn
- Create a virtual environment using
python -m venv faster-rcnn
- Activate virtual environment with
.\faster-rcnn\Scripts\activate
in a Windows OS or withsource ./faster-rcnn/bin/activate
in a Unix OS - Upgrade
pip
usingpython -m pip install --upgrade pip
- Install Cython using
python -m pip install Cython
- Install requirements using
python -m pip install -r requirements.txt
- (Optional) To utilize your CUDA compatible GPU, use
python -m pip install torch==1.10.0+cu102 torchvision==0.11.1+cu102 torchaudio===0.10.0+cu102 -f https://download.pytorch.org/whl/cu102/torch_stable.html
- Setup dataset (format)
- ...
To train Faster R-CNN with custom data, use the training script:
python train.py
You can also test your models after training them using the testing script:
python test.py --model-checkpoint './data/DEMO/model/best.pt' --dataset './data/PennFudanPed/Test'
- TODO: Experiments stats and timing.
All tests were performed using a laptop:
- Processor: Intel(R) Core(TM) i7-9750H CPU @ 2.60GHz 2.59 GHz
- Installed RAM: 16.0 GB (15.85 GB usable)
- Graphics card: NVIDIA GeForce GTX 1660 Ti