A computer vision network for brand logo identification based on the SimSiam neural network. This network was created to satisfy the CSE 586 Computer Vision II class project requirement.
This PyTorch implementation is based on the SimSiam paper by Chen, et al.
- Car Brand Logos Kaggle dataset - a dataset containing 2,913 images of 8 car logos. Each brand has 300~350 training photos and ~50 test photos.
- FlickerLogos Dataset - a brand logos dataset that contains 8,240 images of 32 logo classes from different industries.
Python simsiam_logo_training.py location-of-new-dataset --resume location-of-pretrained-network-checkpoint
Python simsiam_flickr_training.py --resume location-of-pretrained-network-checkpoint
Python simsiam_logo_classification.py location-of-new-dataset --pretrained location-of-pretrained-network-checkpoint
Python simsiam_flickr_classification.py --pretrained location-of-pretrained-network-checkpoint
This project is under the CC-BY-NC 4.0 license (same as the SimSiam paper code).
Dylan Knowles and Akash Kumar