Create app.py

This Python code creates an interactive image classification application using AutoGluon MultiModal and Gradio to determine if a picture contains a tomato.

Key Functionality and Components
The application performs the following main steps:

Model Setup: It uses the huggingface_hub library to download and extract a pre-trained AutoGluon MultiModal image predictor from a specified Hugging Face model repository (apsora/autoML_images_data). This model is loaded locally.

Prediction Logic: The do_predict function takes an uploaded image, saves it temporarily, and then uses the loaded MultiModalPredictor to classify the image into one of two classes: "🍅 Tomato" or "🚫 Not a tomato." It returns the class probabilities for display.

Interactive User Interface (Gradio):
It creates a user-friendly web interface using Gradio. Users can upload an image or capture one using a webcam.When a new image is provided, the do_predict function runs automatically and the result is displayed in a Gradio Label component, showing the predicted class and the confidence score (probability) for both "Tomato" and "Not a tomato." Example images are provided to demonstrate the application's capabilities.

In essence, this is a deployable minimal example demonstrating how to serve a machine learning model, specifically an AutoGluon image classifier, within a Gradio interface.

app.py

@@ -0,0 +1,124 @@
+!pip install autogluon.multimodal --quiet
+
+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
+
+# Hardcoded Hub model (native zip)
+MODEL_REPO_ID = "apsora/autoML_images_data"
+ZIP_FILENAME  = "autogluon_image_predictor_dir.zip"
+HF_TOKEN = os.getenv("HF_TOKEN", None)
+
+# 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",
+        token=HF_TOKEN,
+        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)
+
+# Explicit class labels (edit copy as desired)
+CLASS_LABELS = {0: "🚫 Not a tomato", 1: "🍅 Tomato"}
+
+# Helper to map model class -> human label
+def _human_label(c):
+    try:
+        ci = int(c)
+        return CLASS_LABELS.get(ci, str(c))
+    except Exception:
+        return CLASS_LABELS.get(c, str(c))
+
+# Do the prediction!
+def do_predict(pil_img: PIL.Image.Image):
+    # Make sure there's actually an image to work with
+    if pil_img is None:
+        return "No image provided.", {}, pandas.DataFrame(columns=["Predicted label", "Confidence (%)"])
+
+    # IF we have something to work with, save it and prepare the input
+    tmpdir = pathlib.Path(tempfile.mkdtemp())
+    img_path = tmpdir / "input.png"
+    pil_img.save(img_path)
+
+    df = pandas.DataFrame({"image": [str(img_path)]})  # For AutoGluon expected input format
+
+    # For class probabilities
+    proba_df = PREDICTOR.predict_proba(df)
+
+    # For user-friendly column names
+    proba_df = proba_df.rename(columns={0: "🚫 Not a tomato (0)", 1: "🍅 Tomato (1)"})
+    row = proba_df.iloc[0]
+
+    # For pretty ranked dict expected by gr.Label
+    pretty_dict = {
+        "🚫 Not a tomato": float(row.get("🚫 Not a tomato (0)", 0.0)),
+        "🍅 Tomato": float(row.get("🍅 Tomato (1)", 0.0)),
+    }
+
+    return pretty_dict
+
+# Representative example images! These can be local or links.
+EXAMPLES = [
+    ["https://dengarden.com/.image/w_1920,q_auto:good,c_limit/MTk3NDQ3MTk3NDE4MDcxMDQ2/how-to-get-the-highest-yield-and-best-flavor-from-tomatoes.jpg"],
+    ["https://www.greenlanedelivery.com/cdn/shop/products/Grapes_White_SL_1200x1200.jpg?v=1671549475"],
+    ["https://agrinigeriaprodsa.blob.core.windows.net/agrifarmer/a8738a87-3e02-4d1c-8ba7-e028205ee6bb.jpg"]
+]
+
+# Gradio UI
+with gradio.Blocks() as demo:
+
+    # Provide an introduction
+    gradio.Markdown("# Tomato or No Tomato?")
+    gradio.Markdown("""
+    This is a simple app that demonstrates how to use an autogluon multimodal
+    predictor in a gradio space to predict the contents of a picture. To use,
+    just upload a photo. The result should be generated automatically.
+    """)
+
+    # Interface for the incoming image
+    image_in = gradio.Image(type="pil", label="Input image", sources=["upload", "webcam"])
+
+    # Interface elements to show htte result and probabilities
+    proba_pretty = gradio.Label(num_top_classes=2, label="Class probabilities")
+
+    # Whenever a new image is uploaded, update the result
+    image_in.change(fn=do_predict, inputs=[image_in], outputs=[proba_pretty])
+
+      # For clickable example images
+    gradio.Examples(
+        examples=EXAMPLES,
+        inputs=[image_in],
+        label="Representative examples",
+        examples_per_page=8,
+        cache_examples=False,
+    )
+
+if __name__ == "__main__":
+    demo.launch()