Upload app.py

app.py

@@ -0,0 +1,195 @@
+"""
+🏃 Human Activity Recognition — Gradio Demo
+Fine-tuned MobileNetV2 classifying 15 human activities from images.
+Model: Rishi2455/Human-Activity-Recognition
+"""
+
+import os
+import gradio as gr
+import tensorflow as tf
+import numpy as np
+from PIL import Image
+from huggingface_hub import hf_hub_download
+from datasets import load_dataset
+
+# ── Configuration ────────────────────────────────────────────────────────────
+MODEL_REPO = "Rishi2455/Human-Activity-Recognition"
+MODEL_FILE = "mobilenetv2_finetuned.h5"
+IMG_SIZE = (224, 224)
+
+CLASS_NAMES = [
+    "Calling", "Clapping", "Cycling", "Dancing", "Drinking",
+    "Eating", "Fighting", "Hugging", "Laughing", "Listening to Music",
+    "Running", "Sitting", "Sleeping", "Texting", "Using Laptop",
+]
+
+ACTIVITY_EMOJI = {
+    "Calling": "📞", "Clapping": "👏", "Cycling": "🚴", "Dancing": "💃",
+    "Drinking": "🥤", "Eating": "🍽️", "Fighting": "🥊", "Hugging": "🤗",
+    "Laughing": "😂", "Listening to Music": "🎧", "Running": "🏃",
+    "Sitting": "🪑", "Sleeping": "😴", "Texting": "📱", "Using Laptop": "💻",
+}
+
+# ── Download & load model ───────────────────────────────────────────────────
+print("⬇️  Downloading model...")
+model_path = hf_hub_download(repo_id=MODEL_REPO, filename=MODEL_FILE)
+print("🔧  Loading model...")
+model = tf.keras.models.load_model(model_path, compile=False)
+print("✅  Model loaded!")
+
+# ── Prepare example images from dataset ─────────────────────────────────────
+EXAMPLE_DIR = "examples"
+os.makedirs(EXAMPLE_DIR, exist_ok=True)
+
+example_paths = []
+try:
+    print("📦  Loading example images from dataset...")
+    ds = load_dataset("Bingsu/Human_Action_Recognition", split="test")
+    # Pick one example per class (first 8 classes for a clean grid)
+    selected_classes = [2, 3, 10, 11, 5, 14, 0, 8]  # cycling, dancing, running, sitting, eating, laptop, calling, laughing
+    class_names_lower = [
+        "calling", "clapping", "cycling", "dancing", "drinking",
+        "eating", "fighting", "hugging", "laughing", "listening_to_music",
+        "running", "sitting", "sleeping", "texting", "using_laptop",
+    ]
+    seen = set()
+    for row in ds:
+        label = row["labels"]
+        if label in selected_classes and label not in seen:
+            img = row["image"]
+            fname = f"{EXAMPLE_DIR}/{class_names_lower[label]}.jpg"
+            img.save(fname)
+            example_paths.append(fname)
+            seen.add(label)
+        if len(seen) == len(selected_classes):
+            break
+    print(f"✅  Saved {len(example_paths)} example images.")
+except Exception as e:
+    print(f"⚠️  Could not load examples: {e}")
+
+# ── Inference ────────────────────────────────────────────────────────────────
+def predict(pil_img: Image.Image) -> dict:
+    """Classify a human activity from an image."""
+    if pil_img is None:
+        return {}
+    img = pil_img.convert("RGB").resize(IMG_SIZE)
+    arr = np.expand_dims(np.array(img, dtype=np.float32), axis=0)
+    arr = tf.keras.applications.mobilenet_v2.preprocess_input(arr)
+    preds = model.predict(arr, verbose=0)[0]
+    emoji_labels = {
+        f"{ACTIVITY_EMOJI.get(c, '')} {c}": float(preds[i])
+        for i, c in enumerate(CLASS_NAMES)
+    }
+    return emoji_labels
+
+# ── Gradio UI ────────────────────────────────────────────────────────────────
+DESCRIPTION = """
+Upload a photo of a person performing an activity, and the model will predict which of **15 activities** they are doing.
+
+**Supported activities:** Calling · Clapping · Cycling · Dancing · Drinking · Eating · Fighting · Hugging · Laughing · Listening to Music · Running · Sitting · Sleeping · Texting · Using Laptop
+
+**Model:** [MobileNetV2](https://huggingface.co/Rishi2455/Human-Activity-Recognition) fine-tuned on the [Human Action Recognition dataset](https://huggingface.co/datasets/Bingsu/Human_Action_Recognition)
+"""
+
+css = """
+.main-header { text-align: center; margin-bottom: 0.5rem; }
+.main-header h1 { font-size: 2.2rem; margin-bottom: 0; }
+.footer { text-align: center; margin-top: 1rem; color: #888; font-size: 0.85rem; }
+"""
+
+with gr.Blocks(
+    theme=gr.themes.Soft(
+        primary_hue="blue",
+        secondary_hue="sky",
+        font=gr.themes.GoogleFont("Inter"),
+    ),
+    css=css,
+    title="🏃 Human Activity Recognition",
+    analytics_enabled=False,
+) as demo:
+
+    # Header
+    gr.HTML("""
+    <div class="main-header">
+        <h1>🏃 Human Activity Recognition</h1>
+        <p style="color: #555; font-size: 1.1rem;">Powered by MobileNetV2 · 15 Activity Classes</p>
+    </div>
+    """)
+
+    gr.Markdown(DESCRIPTION)
+
+    with gr.Row(equal_height=True):
+        with gr.Column(scale=1):
+            image_input = gr.Image(
+                type="pil",
+                label="📸 Upload Image",
+                sources=["upload", "webcam", "clipboard"],
+                height=380,
+            )
+            with gr.Row():
+                clear_btn = gr.ClearButton(
+                    [image_input],
+                    value="🗑️ Clear",
+                    variant="secondary",
+                    size="sm",
+                )
+                submit_btn = gr.Button(
+                    "🔍 Classify Activity",
+                    variant="primary",
+                    size="lg",
+                )
+
+        with gr.Column(scale=1):
+            label_output = gr.Label(
+                num_top_classes=5,
+                label="📊 Prediction Results",
+            )
+            gr.Markdown(
+                """
+                <div style="background: #f0f7ff; border-radius: 8px; padding: 12px; margin-top: 8px;">
+                <b>💡 Tips for best results:</b>
+                <ul style="margin: 4px 0;">
+                <li>Use a clear photo with the person visible</li>
+                <li>Single person in the frame works best</li>
+                <li>Good lighting improves accuracy</li>
+                </ul>
+                </div>
+                """
+            )
+
+    # Examples
+    if example_paths:
+        gr.Examples(
+            examples=sorted(example_paths),
+            inputs=image_input,
+            outputs=label_output,
+            fn=predict,
+            cache_examples=True,
+            label="🖼️ Try these examples",
+        )
+
+    # Event handlers
+    submit_btn.click(
+        fn=predict,
+        inputs=image_input,
+        outputs=label_output,
+        api_name="predict",
+    )
+    image_input.change(
+        fn=predict,
+        inputs=image_input,
+        outputs=label_output,
+        api_name=False,
+    )
+
+    # Footer
+    gr.HTML("""
+    <div class="footer">
+        Made with ❤️ using <a href="https://www.gradio.app/" target="_blank">Gradio</a> &
+        <a href="https://huggingface.co/" target="_blank">Hugging Face</a> ·
+        <a href="https://huggingface.co/Rishi2455/Human-Activity-Recognition" target="_blank">Model Card</a>
+    </div>
+    """)
+
+# Launch with show_api=True
+demo.launch(show_api=True)