Proposal: Add Practice Notebook for Zero-Shot Learning Chapter using CLIP

[Medissaoui07]

Hi
I've been going through the Computer Vision course, specifically the chapter on Zero-Shot Learning (ZSL). I found the theoretical explanation very clear and informative.
But u noticed there isn't currently a hands-on code example or notebook for this section. I believe adding one would significantly enhance the learning experience .
I propose adding a new Jupyter notebook to this chapter that demonstrates how to perform zero-shot image classification using OpenAI's CLIP model.

[Medissaoui07]

@mmhamdy i've made already a notebook that covers these steps :
1-Setup: Installing and importing necessary libraries (transformers, datasets, torch, Pillow, matplotlib).
2-Loading: Loading the pre-trained CLIP model/processor and a relevant dataset (e.g., CUB-200-2011, handling class name extraction if needed).
3-ZSL Simulation: Defining a conceptual seen/unseen class split for evaluation purposes.
4-Prompting: Creating appropriate text prompts for the unseen classes.
5-Inference: Running zero-shot classification on images from unseen classes using CLIP.
6-Visualization: Displaying example images with their true and predicted labels to illustrate the results.

[mmhamdy]

Hi @Medissaoui07, that sounds great! Thank you for this amazing effort! 😀 I will have a look at your submitted PR 🙏

[Medissaoui07]

Hi @mmhamdy Thank you! Really, I appreciate your time and feedback. Looking forward to your review, and I’m happy to make any changes if needed!