app.py

# -*- coding: utf-8 -*-
"""Untitled0.ipynb

Automatically generated by Colab.

Original file is located at
    https://colab.research.google.com/drive/1irh48fSmmTfb6dgVI1PbTuzZGFja9htU
"""

import os  # For reading environment variables
import shutil  # For directory cleanup
import zipfile  # For extracting model archives
import pathlib  # For path manipulations
import tempfile  # For creating temporary files/directories

import gradio  # For interactive UI
import pandas  # For tabular data handling
import PIL.Image # For image I/O

import huggingface_hub # For downloading model assets
import autogluon.multimodal  # For loading AutoGluon image classifier

from datasets import load_dataset

# Hardcoded Hub model (native zip)
MODEL_REPO_ID = "SebastianAndreu/2025-24679-HW1-Part2-image-autogluon-predictor"
ZIP_FILENAME  = "autogluon_image_predictor_dir.zip"

# Local cache/extract dirs
CACHE_DIR   = pathlib.Path("hf_assets")
EXTRACT_DIR = CACHE_DIR / "predictor_native"

# Download & load the native predictor
def _prepare_predictor_dir() -> str:
    CACHE_DIR.mkdir(parents=True, exist_ok=True)
    local_zip = huggingface_hub.hf_hub_download(
        repo_id=MODEL_REPO_ID,
        filename=ZIP_FILENAME,
        repo_type="model",
        local_dir=str(CACHE_DIR),
        local_dir_use_symlinks=False,
    )
    if EXTRACT_DIR.exists():
        shutil.rmtree(EXTRACT_DIR)
    EXTRACT_DIR.mkdir(parents=True, exist_ok=True)
    with zipfile.ZipFile(local_zip, "r") as zf:
        zf.extractall(str(EXTRACT_DIR))
    contents = list(EXTRACT_DIR.iterdir())
    predictor_root = contents[0] if (len(contents) == 1 and contents[0].is_dir()) else EXTRACT_DIR
    return str(predictor_root)

PREDICTOR_DIR = _prepare_predictor_dir()
PREDICTOR = autogluon.multimodal.MultiModalPredictor.load(PREDICTOR_DIR)

CLASS_LABELS = {0: "no_duo", 1: "duo"}

def _human_label(c):
    try:
        ci = int(c)
        return CLASS_LABELS.get(ci, str(c))
    except Exception:
        return CLASS_LABELS.get(c, str(c))

# Prediction function
def do_predict(pil_img: PIL.Image.Image):
    if pil_img is None:
        return "No image provided.", {}, pandas.DataFrame(columns=["Predicted label", "Confidence (%)"])

    tmpdir = pathlib.Path(tempfile.mkdtemp())
    img_path = tmpdir / "input.png"
    pil_img.save(img_path)

    df = pandas.DataFrame({"image": [str(img_path)]})
    proba_df = PREDICTOR.predict_proba(df)
    proba_df = proba_df.rename(columns={0: "no_duo (0)", 1: "duo (1)"})
    row = proba_df.iloc[0]

    pretty_dict = {
        "no_duo": float(row.get("no_duo (0)", 0.0)),
        "duo": float(row.get("duo (1)", 0.0)),
    }

    return pretty_dict

# Load the dataset from Hugging Face
# Replace with the correct dataset repo ID you showed in the screenshot
DATASET_REPO_ID = "scottymcgee/duo-image-dataset"
dataset = load_dataset(DATASET_REPO_ID, split="original")

# Save a few example images locally for Gradio's Examples component
tmpdir = pathlib.Path(tempfile.mkdtemp())
EXAMPLES = []

for i in range(5):  # Take the first 5 samples
    img = dataset[i]["image"]       # PIL.Image object
    img_path = tmpdir / f"example_{i}.png"
    img.save(img_path)              # Save to disk
    EXAMPLES.append([str(img_path)])  # Gradio expects file paths in a list

# Define Gradio interface
with gradio.Blocks() as demo:

    # Title and introduction
    gradio.Markdown("# Duo Detection")
    gradio.Markdown("""
    This app demonstrates an **AutoGluon multimodal image predictor**.
    Upload an image, and the model will predict whether it belongs to the class
    `duo` or `no_duo`.
    """)

    # Image input (upload or webcam)
    image_in = gradio.Image(type="pil", label="Input image", sources=["upload", "webcam"])

    # Output: class probabilities
    proba_pretty = gradio.Label(num_top_classes=2, label="Class probabilities")

    # Run prediction whenever a new image is uploaded
    image_in.change(fn=do_predict, inputs=[image_in], outputs=[proba_pretty])

    # Representative examples from the dataset
    gradio.Examples(
        examples=EXAMPLES,
        inputs=[image_in],
        label="Representative examples",
        examples_per_page=5,
        cache_examples=False,
    )

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