Custom Faster R-CNN implementation

• TODO: Introduction to project

Prerequisites

The only prerequisite is git in order to clone the repository. The project was tested using a Python 3.8.5 interpreter.

Installation

1. Open a terminal window
2. Clone repository using git clone https://github.com/AndreasKaratzas/faster-rcnn.git
3. Navigate to project directory with cd faster-rcnn
4. Create a virtual environment using python -m venv faster-rcnn
5. Activate virtual environment with .\faster-rcnn\Scripts\activate in a Windows OS or with source ./faster-rcnn/bin/activate in a Unix OS
6. Upgrade pip using python -m pip install --upgrade pip
7. Install Cython using python -m pip install Cython
8. Install requirements using python -m pip install -r requirements.txt
9. (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

Usage

1. Setup dataset (format)
2. ...

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'

Experiments

• TODO: Experiments stats and timing.

System info

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