This project use Retina Net to do SVHN digit detection.
SVHN is obtained from house numbers in Google Street View images. It contain 10 classes, 1 for each digit. Digit '1' has label 1, '9' has label 9 and '0' has label 10 instead of 0.
torch version 1.7.0
- Download dataset from official site http://ufldl.stanford.edu/housenumbers/
- Clone this repo
- Install Python packages
pip install pandas
pip install pycocotools
pip install opencv-python
pip install requests
Run train.py by following command
python train.py --dataset csv --csv_train <path/to/train_annots.csv> --csv_classes <path/to/train/class_list.csv> --csv_val <path/to/val_annots.csv>
pre-trained model is available at: https://drive.google.com/file/d/1-Szeyz9LA5X_-MvbuYriWovQPEYyrKVI/view?usp=sharing
To visualize image, use:
python inference.py --dataset csv --csv_classes <path/to/train/class_list.csv> --csv_val <path/to/val_annots.csv> --model <path/to/model.pt>