ICVLPR is a research repository focused on machine learning and computer vision, developed as part of an undergraduate thesis project. This repository aims to evaluate the performance of state-of-the-art ANPR models on recognizing commercial vehicle license plates in Indonesia.
The dataset consists of approximately 800 images of commercial vehicle license plates in Indonesia, collected from YouTube videos and direct capture on the streets of Semarang, Indonesia.
The dataset is divided into three sets: train, val, and test, with an 80%, 10%, and 10% split, respectively.
To train the model, make sure you have the following required packages installed:
- PyTorch 2.4.0
- NumPy
- PIL (Python Imaging Library)
- tqdm
- wandb (optional)
or, simply just use the provided requirements.txt using pip:
pip install -r requirements.txtTo start training, run the following command in your shell:
python train.pyYou can also modify the hyperparameters during training. To see the available options, use the --help flag:
python train.py --help