Cardz

Cardz is my take on playing card localization and classification in a scence with multiple cards.

Model

The model consists of an UNet trained with random card ensembles and their segmentation masks. The trained models can be found in results.

Files

This repository contains a notebook explaining the biggest model and analysing its perfomance using Captum.

Here is a basic table of contents:

• Classifier: Training the model
• Captum: Evaluating the model quality using Captum

Dataset

The dataset was generated using several decks of cards and the DTD dataset. @Davanchama (thank you very much) and I made photos of these decks and wrote a script to genrate the dataset. You can find the dataset on kaggle.

References

• [1] U-Net: Convolutional Networks for Biomedical Image Segmentation, Olaf Ronneberger, Philipp Fischer, Thomas Brox
  
• [2] Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization, Ramprasaath R. Selvaraju, Michael Cogswell, et al.

Techstack