shivamlth27/Krishi-ai

Krishi.ai (DFN-PSAN) is an open-source plant disease detection model built on YOLOv5. It replaces YOLOv5’s head with a Pyramid Squeeze Attention (PSA) classification layer and enhances the backbone (CSPDarkNet with a 6×6 conv), SPP pooling, and a DFN neck that fuses FPN and PAN for improved multi-scale feature aggregation. The design emphasizes better feature extraction, attention-guided classification, and real-time detection accuracy. Intended users include researchers, agricultural engineers, extension services, and developers building crop monitoring tools. Key features: modified YOLOv5 backbone, SPP module, DFN neck (FPN+PAN), PSA attention head, multi-scale fusion, PyTorch implementation, MIT license, and a live demo.