README.md

metadata

license: afl-3.0
title: Spaces Readme
sdk: gradio
emoji: 🚀
colorFrom: yellow

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

Lighting DavidNet over Spaces

Gradio Integration for Interactive Model Predictions:

Develop a Gradio interface that envelops the model for seamless interactive predictions. This setup includes the following user-driven functionalities:

GradCAM Visualizations:

• Users can opt to visualize GradCAM images.
• Flexibility to choose the number of GradCAM images for display.
• Selection of the specific layer for generating GradCAM visualizations.
• Ability to adjust opacity of the overlaid GradCAM images.

Misclassified Images:

• Users have the choice to review misclassified images.
• Option to determine the quantity of misclassified images to be presented.

Image Upload:

• Users can upload their own images for predictions.
• A set of 10 example images is available for experimentation.

Class Preferences:

• Users can specify the number of top predicted classes they want to see.
• A limit of 10 classes ensures a manageable display.

Deployment on Hugging Face Spaces:

The Gradio application, featuring the integrated model, is deployed on Hugging Face Spaces. The Spaces README encompasses the following elements:

• A comprehensive overview of the Spaces app's functionality.
• Exclusion of any training-related code from the README.
• Inclusion of links to the Lightning codebase on GitHub, ensuring a clear separation of model training specifics from deployment details.

GitHub Repository for Lightning Training Code:

The Lightning training code resides in a distinct GitHub repository. The repository's detailed README incorporates:

• A comprehensive log detailing the training progression.
• Graphs illustrating the training epochs' loss function.
• Showcasing of 10 instances of misclassified images, complete with actual and predicted labels.

---