Views
No views yet
| Property | Value |
|---|---|
| Task | Image regression |
| Variant | 5deg (5-degree rotation augmentation) |
| Input | RGB image, resized to 224×224 |
| Output | Single scalar (affected area %) |
| Architecture | 3-layer CNN + 2-layer MLP regressor |
1transforms.Compose([
2 transforms.Resize((224, 224)),
3 transforms.ToTensor(),
4])1import torch
2from PIL import Image
3from torchvision import transforms
4
5# Copy model.py from this repo or from the Django app ml/model.py
6from model import GrapeLeafRegressor
7
8model = GrapeLeafRegressor()
9model.load_state_dict(torch.load("grape_leaf_model_5deg.pth", map_location="cpu"))
10model.eval()
11
12transform = transforms.Compose([
13 transforms.Resize((224, 224)),
14 transforms.ToTensor(),
15])
16
17image = Image.open("leaf.jpg").convert("RGB")
18tensor = transform(image).unsqueeze(0)
19
20with torch.no_grad():
21 prediction = model(tensor).item()
22
23prediction = max(0, min(100, prediction))
24print(f"Estimated downy mildew: {prediction:.1f}%")deploy/huggingface/space/ for a Gradio HTTP API used by the Vercel-hosted Django app.data/final_images_5deg (train / validation / test split)python train_model.py --variant 5deg