IOAI-Workshop-CV-1

Multitask learning for Computer Vision; Vision Transformers Basics

Multitask learning for Computer Vision

• Notebook 1: Multitask into (also open in Google Colab)
• Notebook 2: Classification and Segmentation on the Oxford IIIT Pet dataset (also open in Google Colab)
• Bonus Notebook 3: Advanced Data Augmentation techinques and Test Time Augmentation (only Google Colab)

Vision Transformers (ViT) Basics

• Notebook 4: ViT classification benckmark (also open in Google Colab)
• Bonus Notebook 5: ViT - Ensemble methods and Model Soup (also open in Google Colab)

References:

• An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
• Basic ViT implementation
• Advanced ViT implementation
• Pretrained ViTs