from transformers import AutoModelForImageClassification, AutoImageProcessor
from PIL import Image
import torch

# Load model
model = AutoModelForImageClassification.from_pretrained("mesabo/agri-plant-disease-resnet50")
processor = AutoImageProcessor.from_pretrained("mesabo/agri-plant-disease-resnet50")

# Inference
image = Image.open("plant_leaf.jpg").convert("RGB")
inputs = processor(images=image, return_tensors="pt")

with torch.no_grad():
    outputs = model(**inputs)
    probs = torch.nn.functional.softmax(outputs.logits, dim=-1)
    predicted_idx = probs.argmax(-1).item()
    confidence = probs[0][predicted_idx].item()

print(f"Disease: {model.config.id2label[predicted_idx]}")
print(f"Confidence: {confidence * 100:.2f}%")

FastAPI Integration

python

from fastapi import FastAPI, File, UploadFile
from transformers import AutoModelForImageClassification, AutoImageProcessor
from PIL import Image
import torch
import io

app = FastAPI()

# Load model at startup
model = AutoModelForImageClassification.from_pretrained("mesabo/agri-plant-disease-resnet50")
processor = AutoImageProcessor.from_pretrained("mesabo/agri-plant-disease-resnet50")
model.eval()

@app.post("/predict")
async def predict(file: UploadFile = File(...)):
    image = Image.open(io.BytesIO(await file.read())).convert("RGB")
    inputs = processor(images=image, return_tensors="pt")
    
    with torch.no_grad():
        outputs = model(**inputs)
        probs = torch.nn.functional.softmax(outputs.logits, dim=-1)
        predicted_idx = probs.argmax(-1).item()
        confidence = probs[0][predicted_idx].item()
    
    return {
        "disease": model.config.id2label[predicted_idx],
        "confidence": round(confidence * 100, 2),
        "status": "success"
    }

Performance

Metric	Value
Accuracy	95%+
Inference Time	< 100ms (CPU)
Memory Usage	~400 MB
Input Size	224x224 RGB

West African Agriculture Note

This model currently supports Maize and Tomato which are important crops in West Africa.

Not yet supported:

Cassava (most important staple - coming soon)
Cashew (major cash crop)
Cocoa (critical for Ghana, Ivory Coast)

For cassava/cashew support, we recommend fine-tuning on the CCMT Ghana dataset.

Training Data

Dataset: PlantVillage
Images: 54,305
Augmentation: Yes
Resolution: 224x224

Requirements

pip install transformers torch pillow

Citation

bibtex

@misc{agri-plant-disease-resnet50,
  author = {mesabo},
  title = {Plant Disease Detection - ResNet50},
  year = {2024},
  publisher = {Hugging Face},
  url = {https://huggingface.co/mesabo/agri-plant-disease-resnet50}
}

Related Models

mesabo/agri-chat-multilingual - Multilingual agricultural chatbot (7 languages)

License

Apache 2.0 - Commercial use allowed