app.py

import gradio as gr
import torch
import torch.nn.functional as F
from torchvision import models, transforms
from PIL import Image

MODEL_PATH = "densenet121-a639ec97.pth"   # rename your .pth file to model.pth

# ------------------------------------------
# Load Model with Legacy-Key Compatibility
# ------------------------------------------
def load_legacy_densenet(path):
    print("Loading model:", path)
    state_dict = torch.load(path, map_location="cpu")

    model = models.densenet121(weights=None)

    # Legacy checkpoints must use strict=False
    missing, unexpected = model.load_state_dict(state_dict, strict=False)

    print("\n=== LOADING SUMMARY ===")
    print("Missing keys:", len(missing))
    print("Unexpected keys:", len(unexpected))

    model.eval()
    return model

model = load_legacy_densenet(MODEL_PATH)

# ------------------------------------------
# Preprocessing
# ------------------------------------------
transform = transforms.Compose([
    transforms.Resize((224, 224)),
    transforms.ToTensor(),
    transforms.Normalize(
        mean=[0.485, 0.456, 0.406],
        std=[0.229, 0.224, 0.225]
    )
])

# ------------------------------------------
# Prediction function
# ------------------------------------------
def predict(image):
    image = Image.fromarray(image).convert("RGB")
    img_tensor = transform(image).unsqueeze(0)

    with torch.no_grad():
        logits = model(img_tensor)
        probs = F.softmax(logits, dim=1)
        top5_prob, top5_idx = torch.topk(probs, 5)

    results = []
    for p, idx in zip(top5_prob[0], top5_idx[0]):
        results.append({
            "class_index": int(idx.item()),
            "confidence": float(p.item())
        })

    return results

# ------------------------------------------
# Gradio UI
# ------------------------------------------
demo = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="numpy", label="Upload Image"),
    outputs=gr.JSON(label="Top-5 Predictions (Class Index Only)"),
    title="DenseNet-121 Model Classifier (No Label File)",
    description="This app uses a legacy DenseNet121 .pth model and returns top-5 class indices + confidence.",
)

demo.launch()