init

README.md

@@ -1,14 +1,17 @@
 ---
 title: Hand Visibility Detector
-emoji: 😻
-colorFrom: red
+emoji: 🤚
+colorFrom: green
 colorTo: indigo
 sdk: gradio
 sdk_version: 6.12.0
+python_version: 3.12
 app_file: app.py
 pinned: false
 license: cc-by-nc-4.0
 short_description: Per-keypoint Hand Visibility Detector
+models:
+  - ryhara/hand_visibility_detector
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,127 @@
+"""Gradio demo: interactive hand visibility detection on images and videos."""
+
+from __future__ import annotations
+
+import tempfile
+
+import cv2
+import gradio as gr
+import numpy as np
+import torch
+
+from hand_visibility_detector import HandVisibilityPipeline, draw_detections
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+pipe = HandVisibilityPipeline(
+    device=device,
+    dtype=torch.float32,
+)
+
+
+def process_image(
+    image_rgb: np.ndarray,
+    hand_conf: float,
+    show_bones: bool,
+) -> tuple[np.ndarray, str]:
+    if image_rgb is None:
+        return np.zeros((256, 256, 3), dtype=np.uint8), "No image provided"
+    pipe.hand_conf = hand_conf
+    results = pipe.predict(image_rgb)
+    annotated = draw_detections(image_rgb, results, show_bones=show_bones)
+    info_lines = [f"Detected {len(results)} hand(s)"]
+    for i, r in enumerate(results):
+        side = "R" if r.is_right else "L"
+        avg_vis = r.visibility.mean()
+        info_lines.append(
+            f"  [{i}] {side}  conf={r.bbox_conf:.2f}  avg_vis={avg_vis:.2f}"
+        )
+    return annotated, "\n".join(info_lines)
+
+
+def process_video(
+    video_path: str,
+    hand_conf: float,
+    show_bones: bool,
+    progress: gr.Progress = gr.Progress(),
+) -> str | None:
+    if video_path is None:
+        return None
+    import imageio
+
+    cap = cv2.VideoCapture(video_path)
+    if not cap.isOpened():
+        return None
+    fps = cap.get(cv2.CAP_PROP_FPS) or 30.0
+    total = int(cap.get(cv2.CAP_PROP_FRAME_COUNT)) or 0
+
+    out_path = tempfile.mktemp(suffix=".mp4")
+    pipe.hand_conf = hand_conf
+
+    progress(0.0, desc="Starting...")
+    with imageio.get_writer(
+        out_path,
+        fps=fps,
+        codec="libx264",
+        pixelformat="yuv420p",
+        macro_block_size=1,
+        ffmpeg_params=["-movflags", "+faststart"],
+    ) as writer:
+        frame_idx = 0
+        while True:
+            ret, frame = cap.read()
+            if not ret:
+                break
+            frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+            results = pipe.predict(frame_rgb)
+            annotated = draw_detections(frame_rgb, results, show_bones=show_bones)
+            writer.append_data(annotated)
+            frame_idx += 1
+            if total > 0:
+                progress(frame_idx / total, desc=f"Frame {frame_idx}/{total}")
+            else:
+                progress(0.5, desc=f"Frame {frame_idx}")
+
+    cap.release()
+    progress(1.0, desc="Done")
+    return out_path
+
+
+with gr.Blocks(title="Hand Visibility Detector") as demo:
+    gr.Markdown("## Hand Visibility Detector")
+    gr.Markdown(
+        "Detect hands, estimate 3D pose (WiLoR-mini), and predict "
+        "per-keypoint visibility. Green = visible, Red = occluded."
+    )
+
+    with gr.Row():
+        hand_conf_slider = gr.Slider(
+            minimum=0.1, maximum=0.9, value=0.5, step=0.05,
+            label="Hand detection confidence",
+        )
+        show_bones_cb = gr.Checkbox(value=True, label="Show bones")
+
+    with gr.Tabs():
+        with gr.Tab("Single Image"):
+            with gr.Row():
+                img_input = gr.Image(label="Input", type="numpy")
+                img_output = gr.Image(label="Result", type="numpy")
+            img_info = gr.Textbox(label="Info", interactive=False)
+            img_btn = gr.Button("Detect")
+            img_btn.click(
+                fn=process_image,
+                inputs=[img_input, hand_conf_slider, show_bones_cb],
+                outputs=[img_output, img_info],
+            )
+
+        with gr.Tab("Video"):
+            vid_input = gr.Video(label="Input video")
+            vid_output = gr.Video(label="Result video")
+            vid_btn = gr.Button("Process Video")
+            vid_btn.click(
+                fn=process_video,
+                inputs=[vid_input, hand_conf_slider, show_bones_cb],
+                outputs=[vid_output],
+            )
+
+demo.launch()

packages.txt

@@ -0,0 +1,2 @@
+libgl1
+libglib2.0-0

requirements.txt

@@ -0,0 +1 @@
+hand-visibility-detector[demo] @ git+https://github.com/ryhara/hand_visibility_detector.git