This is an implementation of LOGOS on Python 3.7.13, IceVision, and Fastai. The model generates bounding boxes for each instance of a logo in the image. It's based on retinanet and a ResNet50 backbone.
The Repository includes:
- Jupyter notebook to train the model
- Jupyter notebook to download and annotate the model
- Jupyter notebook to perform inference
- Pretrained weights for the model
- Annotations for the datasets
| File Name | Description | |
|---|---|---|
| inference.ipynb | It is the easiest way to start. It shows an example of how to perform inference on an image with the help of the pre trained weights. |  |
| download_and_annotate.ipynb | It shows how to download and generate the annotations for LogoDet3K and Visually29K datasets | |
| Training.ipynb | Shows how to train the model using LogoDet-3K and Visually29K datasets | |
| Dataset | Link |
|---|---|
| LogoDet-3K | https://github.com/Wangjing1551/LogoDet-3K-Dataset |
| Visuallydata | https://github.com/diviz-mit/visuallydata |
| Weights | Link |
|---|---|
| logo-retinanet-checkpoint-52k_384_50.pth | https://drive.google.com/file/d/1GKL15g_-g3xpUwW8tVLMKQPIv-7C77ql/view?usp=sharing |
| logo-retinanet-checkpoint-30000_30.pth | https://drive.google.com/file/d/1iLAgefVtpRZhMZwCrO1hfFpBXq5CDcrS/view?usp=sharing |
Contributions to this repository are welcome.
Python 3.7, torch==1.10.0, torchvision==0.11.1, icevision==0.12.0
If you are using colab all installation instructions are given in the first cell of the notebooks