Update app.py

app.py

@@ -4,21 +4,21 @@ import zipfile  # For extracting model archives
 import pathlib  # For path manipulations
 import tempfile  # For creating temporary files/directories
 
-import gradio as gr  # For interactive UI
-import pandas as pd  # For tabular data handling
-from PIL import Image  # For image I/O
+import gradio  # For interactive UI
+import pandas  # For tabular data handling
+import PIL # For image I/O
 
 import huggingface_hub as hf  # For downloading model assets
 from autogluon.multimodal import MultiModalPredictor  # For loading AutoGluon image classifier
 
 # Hardcoded Hub model (native zip)
-MODEL_REPO_ID = "ccm/2025-24679-image-autogluon-predictor"  # For pointing to your HF repo
-ZIP_FILENAME  = "autogluon_image_predictor_dir.zip"  # For specifying the zipped predictor dir
-HF_TOKEN = os.getenv("HF_TOKEN", None)  # For private repos (optional)
+MODEL_REPO_ID = "ccm/2025-24679-image-autogluon-predictor"
+ZIP_FILENAME  = "autogluon_image_predictor_dir.zip"
+HF_TOKEN = os.getenv("HF_TOKEN", None) 
 
 # Local cache/extract dirs
-CACHE_DIR   = pathlib.Path("hf_assets")  # For caching downloaded assets
-EXTRACT_DIR = CACHE_DIR / "predictor_native"  # For extracted predictor directory
+CACHE_DIR   = pathlib.Path("hf_assets")
+EXTRACT_DIR = CACHE_DIR / "predictor_native"
 
 # Download & load the native predictor
 def _prepare_predictor_dir() -> str:  # For ensuring predictor directory is ready
@@ -40,32 +40,32 @@ def _prepare_predictor_dir() -> str:  # For ensuring predictor directory is read
     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()  # For path to predictor root
-PREDICTOR = MultiModalPredictor.load(PREDICTOR_DIR)  # For loading the image classifier
+PREDICTOR_DIR = _prepare_predictor_dir()
+PREDICTOR = MultiModalPredictor.load(PREDICTOR_DIR)  
 
 # Explicit class labels (edit copy as desired)
-CLASS_LABELS = {0: "♻️ Recycling", 1: "🗑️ Trash"}  # For mapping numeric predictions to human labels
+CLASS_LABELS = {0: "♻️ Recycling", 1: "🗑️ Trash"}  
 
 # Helper to map model class -> human label
-def _human_label(c):  # For robust label mapping
+def _human_label(c): 
     try:
         ci = int(c)
         return CLASS_LABELS.get(ci, str(c))
     except Exception:
         return CLASS_LABELS.get(c, str(c))
 
-# Inference
-def do_predict(pil_img: Image.Image):  # For running a single-image prediction
+# 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.", {}, pd.DataFrame(columns=["Predicted label", "Confidence (%)"])
-    tmpdir = pathlib.Path(tempfile.mkdtemp())  # For a temporary working directory
-    img_path = tmpdir / "input.png"  # For a temporary image path
-    pil_img.save(img_path)
+        return "No image provided.", {}, pandas.DataFrame(columns=["Predicted label", "Confidence (%)"])
 
-    df = pd.DataFrame({"image": [str(img_path)]})  # For AutoGluon expected input format
+    # 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)
 
-    y_pred = PREDICTOR.predict(df)  # For predicted class ids
-    pred_label = _human_label(y_pred.iloc[0])  # For human-readable predicted label
+    df = pandas.DataFrame({"image": [str(img_path)]})  # For AutoGluon expected input format
 
     try:
         proba_df = PREDICTOR.predict_proba(df)  # For class probabilities
@@ -78,49 +78,31 @@ def do_predict(pil_img: Image.Image):  # For running a single-image prediction
             "🗑️ Trash": float(row.get("🗑️ Trash (1)", 0.0)),
         }
         pretty_dict = dict(sorted(pretty_dict.items(), key=lambda kv: kv[1], reverse=True))
-        confidence_pct = round(pretty_dict.get(pred_label.replace(" (0)", "").replace(" (1)", ""), 0.0) * 100, 2)
     except Exception:
-        proba_df = None
         pretty_dict = {}
-        confidence_pct = 100.0  # For default when probabilities are unavailable
 
-    md = f"**Prediction:** {pred_label}"  # For concise summary line
-    if pretty_dict:
-        md += f"  \n**Confidence:** {confidence_pct}%"
-
-    compact = pd.DataFrame([{"Predicted label": pred_label, "Confidence (%)": confidence_pct}])  # For compact table
-    return md, pretty_dict, (proba_df if proba_df is not None else pd.DataFrame())
+    return pretty_dict
 
 # Representative example images (replace with your own)
 EXAMPLES = [
-    ["https://c8.alamy.com/comp/2AEA4K9/a-garbage-and-recycling-can-on-the-campus-of-carnegie-mellon-university-pittsburgh-pennsylvania-usa-2AEA4K9.jpg"],  # For campus bins example
-    ["https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcSvid9M7DynMcoUsX0KBMxooLvrKQJwREiw6g&s"],  # For generic bin example
+    ["https://c8.alamy.com/comp/2AEA4K9/a-garbage-and-recycling-can-on-the-campus-of-carnegie-mellon-university-pittsburgh-pennsylvania-usa-2AEA4K9.jpg"], 
+    ["https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcSvid9M7DynMcoUsX0KBMxooLvrKQJwREiw6g&s"],  
 ]
 
 # Gradio UI
-with gr.Blocks(
-    title="AutoGluon Image Classification — Recycling vs Trash",  # For browser tab title
-    css=".gradio-container {max-width: 900px !important;}"  # For a slightly wider layout
-) as demo:
-    gr.Markdown(  # For top-of-app instructions
-        "## AutoGluon Image Classification — Recycling vs Trash\n"
-        f"**Model:** `{MODEL_REPO_ID}/{ZIP_FILENAME}`  \n"
-        "Drop an image and the prediction updates automatically.\n\n"
-        "- **Class 0 → ♻️ Recycling**\n"
-        "- **Class 1 → 🗑️ Trash**"
-    )
+with gradio.Blocks() as demo:
 
-    image_in = gr.Image(type="pil", label="Input image", sources=["upload", "clipboard", "webcam"])  # For image input
+    # Interface for the incoming image
+    image_in = gradio.Image(type="pil", label="Input image", sources=["upload", "webcam"])  
 
-    pred_md = gr.Markdown()  # For concise summary
-    proba_pretty = gr.Label(num_top_classes=2, label="Class probabilities")  # For pretty ranked probs
-    proba_full = gr.Dataframe(label="Class probabilities (table)", interactive=False)  # For full table view
-    compact = gr.Dataframe(label="Prediction (compact)", interactive=False)  # For 1-row summary table
+    # Interface elements to show htte result and probabilities
+    proba_pretty = gradio.Label(num_top_classes=2, label="Class probabilities") 
 
-    image_in.change(fn=do_predict, inputs=[image_in], outputs=[pred_md, proba_pretty, proba_full])  # For live inference
-    image_in.change(fn=do_predict, inputs=[image_in], outputs=[pred_md, proba_pretty, compact])  # For compact table sync
+    # Whenever a new image is uploaded, update the result
+    image_in.change(fn=do_predict, inputs=[image_in], outputs=[proba_pretty]) 
 
-    gr.Examples(  # For clickable example images
+      # For clickable example images
+    gradio.Examples(
         examples=EXAMPLES,
         inputs=[image_in],
         label="Representative examples",
@@ -128,5 +110,5 @@ with gr.Blocks(
         cache_examples=False,
     )
 
-if __name__ == "__main__":  # For launching the app locally
+if __name__ == "__main__": 
     demo.launch(server_name="0.0.0.0", show_api=False)