app.py

import gradio as gr
import os
import torch
from model import create_effnetb2_model
from timeit import default_timer as timer
from typing import Tuple, Dict
import pkg_resources
import logging

# Set up logging
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
logger = logging.getLogger(__name__)

# Check Gradio version
try:
    gradio_version = pkg_resources.get_distribution("gradio").version
    logger.info(f"Using Gradio version: {gradio_version}")
except pkg_resources.DistributionNotFound:
    raise ImportError("Gradio is not installed. Please install it using 'pip install gradio'.")

# Load class names
try:
    with open("class_names.txt", "r") as f:
        class_names = [food_name.strip() for food_name in f.readlines()]
    logger.info("Class names loaded successfully")
except FileNotFoundError:
    logger.error("class_names.txt not found")
    raise FileNotFoundError("class_names.txt not found.")

# Model and transforms preparation
try:
    effnetb2, effnetb2_transforms = create_effnetb2_model(num_classes=101)
    logger.info("EfficientNetB2 model created successfully")
except Exception as e:
    logger.error(f"Error creating model: {str(e)}")
    raise Exception(f"Error creating model: {str(e)}")

# Load weights
try:
    effnetb2.load_state_dict(
        torch.load(
            "09_pretrained_effnetb2_feature_extractor_food101.pth",
            map_location=torch.device("cpu"),
        )
    )
    logger.info("Model weights loaded successfully")
except FileNotFoundError:
    logger.error("Model weights file not found")
    raise FileNotFoundError("Model weights file not found.")
except Exception as e:
    logger.error(f"Error loading weights: {str(e)}")
    raise Exception(f"Error loading weights: {str(e)}")

# Predict function
def predict(img) -> Tuple[Dict, float]:
    try:
        start_time = timer()
        if img is None:
            raise ValueError("Input image is None.")
        img = effnetb2_transforms(img).unsqueeze(0)
        effnetb2.eval()
        with torch.inference_mode():
            pred_probs = torch.softmax(effnetb2(img), dim=1)
        pred_labels_and_probs = {class_names[i]: float(pred_probs[0][i]) for i in range(len(class_names))}
        pred_time = round(timer() - start_time, 5)
        logger.info(f"Prediction completed: {pred_labels_and_probs}, Time: {pred_time}")
        return pred_labels_and_probs, pred_time
    except Exception as e:
        logger.error(f"Prediction failed: {str(e)}")
        return {"error": f"Prediction failed: {str(e)}"}, 0.0

# Gradio app
title = "FoodVision 101 🍔👁"
description = "An EfficientNetB2 feature extractor to classify 101 food classes."

try:
    example_list = [["examples/" + example] for example in os.listdir("examples")]
    logger.info("Examples loaded successfully")
except FileNotFoundError:
    example_list = []
    logger.warning("'examples/' directory not found")

# Simplified Gradio interface
demo = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="pil"),
    outputs=[
        gr.Label(num_top_classes=5, label="Predictions"),
        gr.Number(label="Prediction time (s)"),
    ],
    examples=example_list,
    title=title,
    description=description,
    allow_flagging="never",  # Disable flagging to simplify API
    api_mode=False,          # Disable API mode to avoid schema generation
)

# Launch with share=True for Hugging Face Spaces
try:
    demo.launch(share=True)
    logger.info("Gradio app launched successfully")
except Exception as e:
    logger.error(f"Failed to launch Gradio app: {str(e)}")
    raise Exception(f"Failed to launch Gradio app: {str(e)}")