Code for AI Ukraine 2018 workshop "Introduction to Capsule Networks"
Please note! Source code for workshop can be changed up to Friday, 11 Oct.(freeze date). Clone the latest version please.
-
To prepare for the workshop, please:
- Download Python notebook.
- Register/try Google Colab https://colab.research.google.com/notebook
- Use notebook in Google Colab (preffered way) or install dependencies and run notebook locally.
-
Dataset used: German Traffic Sign Dataset. http://benchmark.ini.rub.de/?section=gtsrb&subsection=dataset This dataset has more than 50,000 images of 43 classes. You can compare other models with other implementations.
Pickled, prepared dataset you can find here: https://d17h27t6h515a5.cloudfront.net/topher/2017/February/5898cd6f_traffic-signs-data/traffic-signs-data.zip (This scaled and pickled dataset was taken from the source: https://github.com/mohamedameen93/German-Traffic-Sign-Classification-Using-TensorFlow , you can find LeNet-5 and VGGNet models implementation for this dataset)
- Final version of the notebook with source will be presented during the workshop.