Logo Lanfrica

SAIFULLAH-SHARAFAT/A-Self-Supervised-Deep-Learning-Framework-for-Malabar-Spinach-Leaf-Disease-Classification

Domaine:

agriculture

Type de record:

modeldataset
Créateur:
SAI
HĂ´te:
self-supervised learning for low-resource crops # 🌿 Malabar Spinach Leaf Disease Classification *A Self-Supervised Attention Framework with Vision Transformers* --- ## 🎥 Web Demo Video --- ## 🔗 HD Preview (LinkedIn) --- # 🔗 Resources ### 🧠 Pretrained Models (HuggingFace) huggingface.co * `simsiam_cbam_pretrained_final.pth` → Self-supervised SimSiam backbone * `best_finetuned_cbam.pth` → **Final deployment model (SimSiam-CBAM-ResNet-50, CE)** * Additional experimental checkpoints (`*.ckpt`) available ### 🗂 Dataset (3 Classes) huggingface.co * **Raw images:** ~700+ * **Augmented total:** ~2100+ * **Classes:** Alternaria • Straw Mite • Healthy * Split: 70% train / 15% val / 15% test * Collected and verified by agricultural experts (Habiganj Agricultural University) ### 📚 Code Notebooks | Model / Experiment | Notebook | | ------------------------------------------------------ | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **SimSiam + CBAM (final model)** | simsiam-cbam-plos-rev.ipynb | | EfficientNetB0, SpinachCNN, Spinach-ResSENet, ViT-B/16 | spinachleaf.ipynb | | Vanilla SimSiam (ResNet-50) | vanilla-resnet.ipynb | | SwinV2-Base (scratch + pretrained) | swin-spinach.ipynb | --- # 🧭 Project Overview This repository presents a **domain-optimized, interpretable, and deployment-friendly framework** for Malabar spinach leaf disease diagnosis, integrating: # …