iamgkkj/LeafGuard-backend

LeafGuard AI

Multi-class crop leaf disease identification with CBAM + MobileNetV3-Small and Grad-CAM explainability.

Upload a leaf image and choose which model(s) to run:

• Compare both (default) — Shows predictions from both Baseline and CBAM models
• CBAM only — MobileNetV3-Small with Convolutional Block Attention Module
• Baseline only — Standard MobileNetV3-Small

Models

Model	Parameters	Size
Baseline MobileNetV3-Small	1,556,806	6.41 MB
MobileNetV3-Small + CBAM	1,562,524	6.49 MB

Deployment

Backend runs on Hugging Face Spaces (free Gradio SDK tier). Model weights are downloaded at startup from the Hugging Face Hub.

GitHub

https://github.com/iamgkkj/LeafGuard