app.py

import gradio as gr
from PIL import Image
import torch
from torchvision import transforms
from transformers import AutoModelForImageClassification

# Define model names and corresponding labels
MODEL_CONFIGS = [
    {
        "name": "anismizi/skin-type-classifier",
        "labels": ["dry", "oily"],
        "key": "oil_vs_dry"
    },
    {
        "name": "imfarzanansari/skintelligent-acne",      
        "labels": ["no_acne", "acne"],
        "key": "acne"
    },
    {
        "name": "imfarzanansari/skintelligent-wrinkles",
        "labels": ["no_wrinkles", "wrinkles"],
        "key": "wrinkles"
    },
]

# Load all models at startup
MODELS = []
for config in MODEL_CONFIGS:
    model = AutoModelForImageClassification.from_pretrained(config["name"])
    model.eval()
    MODELS.append(model)

# Common preprocessing (adjust if any model requires different input specs)
preprocess = transforms.Compose([
    transforms.Resize(256),
    transforms.CenterCrop(224),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406],
                         std=[0.229, 0.224, 0.225]),
])

def analyze_skin(image: Image.Image):
    image = image.convert("RGB")
    input_tensor = preprocess(image)
    input_batch = input_tensor.unsqueeze(0)  # add batch dimension

    results = {}
    with torch.no_grad():
        for idx, config in enumerate(MODEL_CONFIGS):
            model, labels, key = MODELS[idx], config["labels"], config["key"]
            outputs = model(input_batch)
            logits = outputs.logits
            probs = torch.softmax(logits, dim=1)
            confidence, pred_idx = torch.max(probs, dim=1)
            predicted_label = labels[pred_idx.item()]
            confidence_score = confidence.item()
            results[key] = {
                "label": predicted_label,
                "confidence": f"{confidence_score:.2%}"
            }
    return results

iface = gr.Interface(
    fn=analyze_skin,
    inputs=gr.Image(type="pil"),
    outputs=gr.JSON(label="Skin Analysis Results"),
    title="Comprehensive Skin Condition Analyzer",
    description="Classifies skin image for oily/dry, acne, redness, wrinkles using multiple models."
)

if __name__ == "__main__":
    iface.launch()