Upload folder using huggingface_hub

.dockerignore

@@ -1,5 +1,11 @@
 .git
 .github
+src/
+CMakeLists.txt
+entrypoint.sh
+Cube.stl
+Sphere.stl
+mask_to_coco.py
 build
 build_v2
 __pycache__
@@ -21,4 +27,4 @@ docs/images
 deps/
 build/
 *.dll
-*.a
+*.a

.gitattributes

@@ -1 +1,5 @@
 deps/glfw-3.4.bin.WIN64.zip filter=lfs diff=lfs merge=lfs -text
+demo_data/rgb/frame_0010.png filter=lfs diff=lfs merge=lfs -text
+demo_data/rgb/frame_0025.png filter=lfs diff=lfs merge=lfs -text
+demo_data/rgb/frame_0050.png filter=lfs diff=lfs merge=lfs -text
+demo_data/rgb/frame_0075.png filter=lfs diff=lfs merge=lfs -text

Dockerfile

@@ -1,42 +1,24 @@
-FROM ubuntu:22.04
-
-# 1. 安装 Linux 下渲染和编译所需的所有依赖
-# 安装编译工具和库
-RUN apt-get update && apt-get install -y \
-    cmake \
-    build-essential \
-    pkg-config \
-    libglfw3-dev \
-    libglew-dev \
-    libgl1-mesa-dev \
-    libglu1-mesa-dev \
-    libosmesa6-dev \
-    libbenchmark-dev \
-    libgl-dev \
-    python3-pip \
-    git \
-    && rm -rf /var/lib/apt/lists/*
+FROM python:3.10-slim
 
 WORKDIR /app
 
-# 2. 只拷贝源码和必要文件 (删掉 COPY deps/ 这一行)
-COPY CMakeLists.txt .
-COPY src/ ./src/
-COPY app.py .
 COPY requirements.txt .
-# 如果你有这些脚本和模型文件，保留它们
-COPY mask_to_coco.py .
-COPY Cube.stl .
-COPY Sphere.stl .
+RUN pip install --no-cache-dir -r requirements.txt
+
+COPY app.py .
+COPY demo_data/ ./demo_data/
+COPY demo_showcase.html .
 
-# 3. 编译 C++ 渲染器
-# 关键调试行：如果这里出来的列表没有 .h，说明文件根本没传上来
-RUN ls -R /app/src/core/
+RUN useradd -m -u 1000 user
+RUN chown -R user:user /app
+USER user
 
-# 然后再执行构建
-RUN mkdir build && cd build && cmake .. -DCMAKE_BUILD_TYPE=Release && make -j$(nproc)
-# 4. 安装 Python 依赖
-RUN pip3 install --no-cache-dir -r requirements.txt
+EXPOSE 7860
 
-# 5. 启动
-CMD ["python3", "app.py"]
+ENTRYPOINT ["python3", "-m", "streamlit", "run", "app.py", \
+    "--server.port=7860", \
+    "--server.address=0.0.0.0", \
+    "--server.headless=true", \
+    "--server.enableCORS=false", \
+    "--server.enableXsrfProtection=false", \
+    "--browser.gatherUsageStats=false"]

README.md

@@ -1,8 +1,18 @@
----
-title: Huhb3D Synthetic Data Generator
-emoji: 🤖
-colorFrom: blue
-colorTo: indigo
-sdk: docker
-pinned: false
----
+---
+title: Huhb3D Synthetic Data Generator
+emoji: 🤖
+colorFrom: blue
+colorTo: indigo
+sdk: docker
+pinned: false
+---
+
+# Huhb3D Synthetic Data Generator (Demo)
+
+面向机器人视觉训练的合成数据生成器 — 在线 Demo 展示
+
+⚠️ 此在线版本为 **Demo 模式**，展示预生成的样例数据（RGB + 语义Mask + 6DoF位姿 + BOP格式）。
+
+要使用自己的 CAD 模型生成数据，请在本地部署：
+- GitHub: https://github.com/AIminminAI/Huhb3D-Viewer
+- Gitee: https://gitee.com/aiminminai/Huhb3D-Viewer

app.py

@@ -1,716 +1,221 @@
-"""
-app.py - Huhb3D Synthetic Data Generator Web UI
-================================================
-Streamlit-based web interface for the synthetic data pipeline.
-
-Usage:
-    streamlit run app.py
-"""
-
-import streamlit as st
-import subprocess
-import sys
-import os
-import shutil
-import zipfile
-import json
-import time
-import tempfile
-import re
-import random
-import base64
-from pathlib import Path
-
-import requests
-
-
-SUPPORTED_EXTENSIONS = ["step", "stp", "iges", "igs", "stl", "obj"]
-SUPPORTED_UPLOAD_TYPES = [".step", ".stp", ".iges", ".igs", ".stl", ".obj"]
-
-SCRIPT_DIR = Path(__file__).parent.resolve()
-TEMP_DIR = SCRIPT_DIR / "temp_uploads"
-OUTPUT_DIR = SCRIPT_DIR / "streamlit_output"
-
-
-DEEPSEEK_API_URL = "https://api.deepseek.com/v1/chat/completions"
-DEEPSEEK_PROMPT = (
-    "你是一个工业质检专家，请描述这张3D零件图中的关键特征"
-    "（如：法兰盘、沉头孔、中心轴等），用于训练机器人视觉模型。"
-    "请按以下JSON格式输出：\n"
-    '{"part_type": "零件类型", "key_features": ["特征1", "特征2", ...], '
-    '"surfaces": ["面类型1", "面类型2", ...], '
-    '"manufacturing_hints": ["加工提示1", ...], '
-    '"description": "详细描述"}'
-)
-
-
-def call_deepseek_vision(image_path, api_key, model="deepseek-chat"):
-    try:
-        with open(image_path, "rb") as f:
-            img_b64 = base64.b64encode(f.read()).decode("utf-8")
-
-        ext = Path(image_path).suffix.lower()
-        mime_map = {".png": "image/png", ".jpg": "image/jpeg", ".jpeg": "image/jpeg"}
-        mime_type = mime_map.get(ext, "image/png")
-
-        payload = {
-            "model": model,
-            "messages": [
-                {
-                    "role": "user",
-                    "content": [
-                        {"type": "text", "text": DEEPSEEK_PROMPT},
-                        {
-                            "type": "image_url",
-                            "image_url": {
-                                "url": f"data:{mime_type};base64,{img_b64}"
-                            },
-                        },
-                    ],
-                }
-            ],
-            "max_tokens": 1024,
-            "temperature": 0.3,
-        }
-
-        headers = {
-            "Authorization": f"Bearer {api_key}",
-            "Content-Type": "application/json",
-        }
-
-        resp = requests.post(
-            DEEPSEEK_API_URL,
-            headers=headers,
-            json=payload,
-            timeout=60,
-        )
-        resp.raise_for_status()
-
-        result = resp.json()
-        content = result["choices"][0]["message"]["content"]
-
-        description = {
-            "source_image": Path(image_path).name,
-            "model": model,
-            "prompt": DEEPSEEK_PROMPT,
-            "raw_response": content,
-        }
-
-        try:
-            json_start = content.index("{")
-            json_end = content.rindex("}") + 1
-            parsed = json.loads(content[json_start:json_end])
-            description["parsed"] = parsed
-        except (ValueError, json.JSONDecodeError):
-            description["parsed"] = None
-
-        return True, description
-
-    except requests.exceptions.Timeout:
-        return False, "API request timed out (60s). Please try again."
-    except requests.exceptions.ConnectionError:
-        return False, "Cannot connect to DeepSeek API. Check your network."
-    except requests.exceptions.HTTPError as e:
-        code = e.response.status_code if e.response else "unknown"
-        detail = ""
-        try:
-            detail = e.response.json().get("error", {}).get("message", "")
-        except Exception:
-            pass
-        return False, f"API HTTP error {code}: {detail}"
-    except KeyError as e:
-        return False, f"Unexpected API response format: missing {e}"
-    except Exception as e:
-        return False, f"Unexpected error: {type(e).__name__}: {str(e)}"
-
-
-def find_cpp_executable():
-    search_paths = [
-        SCRIPT_DIR / "build" / "Release" / "test_render.exe",
-        SCRIPT_DIR / "build_v2" / "Release" / "test_render.exe",
-        SCRIPT_DIR / "build" / "test_render.exe",
-        SCRIPT_DIR / "build_v2" / "test_render.exe",
-    ]
-    for path in search_paths:
-        if path.exists():
-            return str(path.resolve())
-    return None
-
-
-def convert_to_stl(input_path, output_stl_path):
-    input_path = Path(input_path)
-    suffix = input_path.suffix.lower()
-
-    if suffix in (".stl",):
-        if str(input_path.resolve()) != str(output_stl_path):
-            shutil.copy2(input_path, output_stl_path)
-        return True
-
-    if suffix in (".step", ".stp", ".iges", ".igs"):
-        try:
-            import cadquery as cq
-            if suffix in (".step", ".stp"):
-                shape = cq.importers.importStep(str(input_path))
-            else:
-                shape = cq.importers.importStep(str(input_path))
-            cq.exporters.export(shape, str(output_stl_path), exportType="STL")
-            return True
-        except ImportError:
-            pass
-        except Exception as e:
-            st.warning(f"cadquery conversion failed: {e}")
-
-        try:
-            import FreeCAD
-            FreeCAD.open(str(input_path))
-            doc = FreeCAD.ActiveDocument
-            if doc is None:
-                st.error("FreeCAD failed to open the file")
-                return False
-            import Mesh
-            mesh_objs = []
-            for obj in doc.Objects:
-                if hasattr(obj, "Shape"):
-                    mesh_data = Mesh.Mesh(obj.Shape.tessellate(0.1))
-                    mesh_objs.append(mesh_data)
-            if mesh_objs:
-                combined = mesh_objs[0]
-                for m in mesh_objs[1:]:
-                    combined = combined.unite(m)
-                combined.write(str(output_stl_path))
-            FreeCAD.closeDocument(doc.Name)
-            return Path(output_stl_path).exists()
-        except ImportError:
-            pass
-        except Exception as e:
-            st.warning(f"FreeCAD conversion failed: {e}")
-
-        st.error(
-            "STEP/IGES format requires a conversion library.\n\n"
-            "Please install one of:\n"
-            "- `pip install cadquery` (recommended)\n"
-            "- Or use FreeCAD with Python bindings\n\n"
-            "Alternatively, convert your file to STL first using any CAD tool."
-        )
-        return False
-
-    if suffix in (".obj",):
-        try:
-            import trimesh
-            mesh = trimesh.load(str(input_path))
-            mesh.export(str(output_stl_path))
-            return True
-        except ImportError:
-            pass
-        except Exception as e:
-            st.warning(f"trimesh conversion failed: {e}")
-
-        st.error(
-            "OBJ format requires `trimesh` for conversion.\n"
-            "Install with: `pip install trimesh`"
-        )
-        return False
-
-    st.error(f"Unsupported format: {suffix}")
-    return False
-
-
-def parse_progress(line, total_count):
-    m = re.search(r'Progress:\s+(\d+)/(\d+)\s+\((\d+\.?\d*)%', line)
-    if m:
-        current = int(m.group(1))
-        total = int(m.group(2))
-        percent = float(m.group(3))
-        return current, total, percent / 100.0
-    return None
-
-
-def run_generation(stl_path, output_dir, sample_count, camera_radius,
-                   image_width, image_height, save_mask, save_depth=False,
-                   progress_callback=None):
-    exe_path = find_cpp_executable()
-    if not exe_path:
-        return False, "C++ executable not found. Please compile the project first."
-
-    cmd = [
-        exe_path,
-        "--batch",
-        "--input", str(stl_path),
-        "--output", str(output_dir),
-        "--count", str(sample_count),
-        "--radius", str(camera_radius),
-        "--width", str(image_width),
-        "--height", str(image_height),
-    ]
-    if not save_mask:
-        cmd.append("--no-mask")
-
-    if save_depth:
-        cmd.append("--depth")
-
-    try:
-        process = subprocess.Popen(
-            cmd,
-            stdout=subprocess.PIPE,
-            stderr=subprocess.STDOUT,
-            text=True,
-            bufsize=1,
-            encoding='utf-8',
-            errors='replace',
-            cwd=str(SCRIPT_DIR),
-        )
-
-        log_lines = []
-        for line in process.stdout:
-            line = line.rstrip()
-            if line:
-                log_lines.append(line)
-                progress_info = parse_progress(line, sample_count)
-                if progress_info:
-                    current, total, pct = progress_info
-                    progress_callback(pct, f"Rendering: {current}/{total}")
-
-        process.wait()
-
-        if process.returncode != 0:
-            return False, "C++ generator failed.\n\nLog:\n" + "\n".join(log_lines[-20:])
-
-        return True, "\n".join(log_lines[-10:])
-
-    except FileNotFoundError:
-        return False, f"Executable not found: {exe_path}"
-    except Exception as e:
-        return False, f"Error: {str(e)}"
-
-
-def package_zip(output_dir):
-    output_path = Path(output_dir)
-    if not output_path.exists():
-        return None
-
-    zip_path = output_path.parent / (output_path.name + ".zip")
-
-    rgb_dir = output_path / "rgb"
-    mask_dir = output_path / "mask"
-    depth_dir = output_path / "depth"
-    legend_file = output_path / "label_legend.txt"
-    desc_file = output_path / "description.json"
-    poses_file = output_path / "camera_poses.json"
-
-    rgb_count = len(list(rgb_dir.glob("*.png"))) if rgb_dir.exists() else 0
-    mask_count = len(list(mask_dir.glob("*.png"))) if mask_dir.exists() else 0
-    depth_count = len(list(depth_dir.glob("*.png"))) if depth_dir.exists() else 0
-
-    with zipfile.ZipFile(zip_path, 'w', zipfile.ZIP_DEFLATED) as zf:
-        if rgb_dir.exists():
-            for f in sorted(rgb_dir.glob("*.png")):
-                zf.write(f, f"rgb/{f.name}")
-        if mask_dir.exists():
-            for f in sorted(mask_dir.glob("*.png")):
-                zf.write(f, f"mask/{f.name}")
-        if depth_dir.exists():
-            for f in sorted(depth_dir.glob("*.png")):
-                zf.write(f, f"depth/{f.name}")
-            for f in sorted(depth_dir.glob("*.raw")):
-                zf.write(f, f"depth/{f.name}")
-        if legend_file.exists():
-            zf.write(legend_file, legend_file.name)
-        if desc_file.exists():
-            zf.write(desc_file, desc_file.name)
-        if poses_file.exists():
-            zf.write(poses_file, poses_file.name)
-        manifest = {
-            "version": "2.0",
-            "generator": "Huhb3D-SyntheticDataPipeline",
-            "rgb_count": rgb_count,
-            "mask_count": mask_count,
-            "depth_count": depth_count,
-            "has_legend": legend_file.exists(),
-            "has_ai_description": desc_file.exists(),
-            "has_camera_poses": poses_file.exists(),
-        }
-        zf.writestr("manifest.json", json.dumps(manifest, indent=2))
-
-    return str(zip_path)
-
-
-def get_generation_stats(output_dir):
-    output_path = Path(output_dir)
-    rgb_dir = output_path / "rgb"
-    mask_dir = output_path / "mask"
-    legend_file = output_path / "label_legend.txt"
-
-    stats = {
-        "rgb_count": len(list(rgb_dir.glob("*.png"))) if rgb_dir.exists() else 0,
-        "mask_count": len(list(mask_dir.glob("*.png"))) if mask_dir.exists() else 0,
-        "has_legend": legend_file.exists(),
-    }
-
-    if legend_file.exists():
-        categories = {}
-        with open(legend_file, 'r') as f:
-            for line in f:
-                line = line.strip()
-                if not line or line.startswith('#'):
-                    continue
-                parts = line.split()
-                if len(parts) >= 4:
-                    cat_id = parts[0]
-                    cat_name = parts[1]
-                    cat_color = parts[2:5]
-                    categories[cat_id] = {"name": cat_name, "color": cat_color}
-        stats["categories"] = categories
-
-    return stats
-
-
-def main():
-    st.set_page_config(
-        page_title="Huhb3D Synthetic Data Generator",
-        page_icon="🤖",
-        layout="wide",
-    )
-
-    st.title("🤖 Huhb3D Synthetic Data Generator")
-    st.markdown("---")
-
-    col_left, col_right = st.columns([1, 2])
-
-    with col_left:
-        st.subheader("📁 Upload CAD Model")
-        uploaded_file = st.file_uploader(
-            "Supports STEP, STP, IGES, IGS, STL, OBJ",
-            type=SUPPORTED_EXTENSIONS,
-            help="Upload your CAD model file. STEP/IGES files will be auto-converted to STL.",
-        )
-
-        if uploaded_file is not None:
-            file_ext = Path(uploaded_file.name).suffix.lower()
-            file_size_mb = uploaded_file.size / (1024 * 1024)
-            st.success(f"✅ {uploaded_file.name} ({file_size_mb:.1f} MB)")
-
-            if file_ext in (".step", ".stp", ".iges", ".igs"):
-                st.info("🔄 STEP/IGES will be auto-converted to STL")
-        else:
-            st.info("👆 Please upload a CAD model file")
-
-        st.markdown("---")
-        st.subheader("⚙️ Generation Settings")
-
-        sample_count = st.slider(
-            "📊 Sample Count",
-            min_value=100,
-            max_value=1000,
-            value=500,
-            step=50,
-            help="Number of 360° views to render",
-        )
-
-        camera_radius = st.slider(
-            "🔭 Camera Radius",
-            min_value=2.0,
-            max_value=20.0,
-            value=5.0,
-            step=0.5,
-            help="Distance of camera from model center",
-        )
-
-        col_w, col_h = st.columns(2)
-        with col_w:
-            image_width = st.selectbox("Width", [512, 640, 800, 1024, 1280], index=2)
-        with col_h:
-            image_height = st.selectbox("Height", [480, 600, 720, 768, 960], index=1)
-
-        save_mask = st.checkbox("🏷️ Generate Semantic Masks", value=True,
-                                help="Generate pixel-level semantic segmentation masks")
-
-        save_depth = st.checkbox("📏 Generate Depth Maps", value=False,
-                                 help="Generate depth maps for robot grasping planning")
-
-        save_camera_poses = st.checkbox("📐 Export Camera Poses (6DoF)", value=True,
-                                        help="Export camera position, rotation, view/projection matrices as JSON")
-
-        st.markdown("---")
-        st.subheader("🧠 AI Description")
-
-        deepseek_api_key = st.text_input(
-            "DeepSeek API Key",
-            type="password",
-            help="Enter your DeepSeek API key to generate AI-powered part descriptions",
-        )
-
-        enable_ai_desc = st.checkbox(
-            "🧠 AI Feature Description",
-            value=False,
-            help="Use DeepSeek-V3 to describe key features of the generated part",
-        )
-
-        if enable_ai_desc and not deepseek_api_key:
-            st.warning("⚠️ Please enter your DeepSeek API Key")
-
-    with col_right:
-        st.subheader("🚀 Generation Pipeline")
-
-        exe_path = find_cpp_executable()
-        if exe_path:
-            st.success(f"✅ C++ Engine: `{Path(exe_path).name}`")
-        else:
-            st.error("❌ C++ Engine not found! Please compile the project first.")
-            st.code("cd build && cmake --build . --config Release", language="bash")
-
-        st.markdown("---")
-
-        start_button = st.button(
-            "🚀 Start Generation",
-            disabled=(uploaded_file is None or exe_path is None),
-            use_container_width=True,
-            type="primary",
-        )
-
-        if start_button and uploaded_file is not None:
-            TEMP_DIR.mkdir(exist_ok=True)
-            OUTPUT_DIR.mkdir(exist_ok=True)
-
-            session_id = str(int(time.time()))
-            session_output = OUTPUT_DIR / f"run_{session_id}"
-
-            file_ext = Path(uploaded_file.name).suffix.lower()
-            temp_input = TEMP_DIR / f"upload_{session_id}{file_ext}"
-            temp_stl = TEMP_DIR / f"upload_{session_id}.stl"
-
-            with open(temp_input, "wb") as f:
-                f.write(uploaded_file.getbuffer())
-
-            st.info(f"📁 File saved: {uploaded_file.name}")
-
-            if file_ext not in (".stl",):
-                with st.spinner("Converting to STL..."):
-                    conv_ok = convert_to_stl(temp_input, temp_stl)
-                if not conv_ok:
-                    st.stop()
-                stl_path = temp_stl
-                st.success("✅ Conversion to STL complete")
-            else:
-                stl_path = temp_input
-
-            progress_bar = st.progress(0.0, text="Initializing...")
-            status_text = st.empty()
-            log_area = st.empty()
-
-            def on_progress(pct, msg):
-                progress_bar.progress(pct, text=msg)
-
-            start_time = time.time()
-
-            success, log_msg = run_generation(
-                stl_path=stl_path,
-                output_dir=session_output,
-                sample_count=sample_count,
-                camera_radius=camera_radius,
-                image_width=image_width,
-                image_height=image_height,
-                save_mask=save_mask,
-                save_depth=save_depth,
-                progress_callback=on_progress,
-            )
-
-            elapsed = time.time() - start_time
-
-            if success:
-                progress_bar.progress(1.0, text="✅ Generation Complete!")
-
-                stats = get_generation_stats(session_output)
-
-                st.markdown("---")
-                st.subheader("📊 Generation Results")
-
-                col_s1, col_s2, col_s3, col_s4 = st.columns(4)
-                with col_s1:
-                    st.metric("RGB Images", stats["rgb_count"])
-                with col_s2:
-                    st.metric("Mask Images", stats["mask_count"])
-                with col_s3:
-                    st.metric("Time", f"{elapsed:.1f}s")
-                with col_s4:
-                    depth_count = len(list((session_output / "depth").glob("*.png"))) if (session_output / "depth").exists() else 0
-                    st.metric("Depth Maps", depth_count)
-
-                if stats.get("categories"):
-                    with st.expander("🏷️ Label Categories", expanded=False):
-                        cat_data = []
-                        for cat_id, cat_info in stats["categories"].items():
-                            r, g, b = cat_info["color"]
-                            cat_data.append({
-                                "ID": cat_id,
-                                "Category": cat_info["name"],
-                                "Color (RGB)": f"({r}, {g}, {b})",
-                            })
-                        st.table(cat_data)
-
-                ai_description = None
-                if enable_ai_desc and deepseek_api_key:
-                    rgb_dir = session_output / "rgb"
-                    if rgb_dir.exists():
-                        png_files = sorted(rgb_dir.glob("*.png"))
-                        if png_files:
-                            sample_img = random.choice(png_files)
-                            st.markdown("---")
-                            st.subheader("🧠 AI Feature Description")
-
-                            col_ai_img, col_ai_text = st.columns([1, 2])
-                            with col_ai_img:
-                                st.image(str(sample_img), caption=f"Sampled: {sample_img.name}",
-                                         use_container_width=True)
-
-                            with col_ai_text:
-                                with st.spinner("Calling DeepSeek-V3 API..."):
-                                    ai_ok, ai_result = call_deepseek_vision(
-                                        str(sample_img), deepseek_api_key
-                                    )
-
-                            if ai_ok:
-                                ai_description = ai_result
-                                desc_path = session_output / "description.json"
-                                with open(desc_path, "w", encoding="utf-8") as f:
-                                    json.dump(ai_description, f, indent=2, ensure_ascii=False)
-
-                                with col_ai_text:
-                                    st.success("✅ AI description generated!")
-
-                                    if ai_description.get("parsed"):
-                                        parsed = ai_description["parsed"]
-                                        st.markdown(f"**Part Type:** {parsed.get('part_type', 'N/A')}")
-
-                                        features = parsed.get("key_features", [])
-                                        if features:
-                                            st.markdown("**Key Features:**")
-                                            for feat in features:
-                                                st.markdown(f"- {feat}")
-
-                                        surfaces = parsed.get("surfaces", [])
-                                        if surfaces:
-                                            st.markdown("**Surfaces:** " + ", ".join(surfaces))
-
-                                        hints = parsed.get("manufacturing_hints", [])
-                                        if hints:
-                                            st.markdown("**Manufacturing Hints:**")
-                                            for h in hints:
-                                                st.markdown(f"- {h}")
-
-                                        desc_text = parsed.get("description", "")
-                                        if desc_text:
-                                            st.markdown(f"**Description:** {desc_text}")
-                                    else:
-                                        st.markdown(ai_description.get("raw_response", ""))
-
-                                    with st.expander("📋 Raw JSON", expanded=False):
-                                        st.json(ai_description)
-                            else:
-                                with col_ai_text:
-                                    st.warning(f"⚠️ AI description failed: {ai_result}")
-                                    st.info("The dataset will be packaged without AI description.")
-                    else:
-                        st.warning("⚠️ No RGB images found for AI description.")
-
-                with st.spinner("Packaging ZIP..."):
-                    zip_path = package_zip(session_output)
-
-                if zip_path:
-                    zip_size_mb = Path(zip_path).stat().st_size / (1024 * 1024)
-
-                    st.markdown("---")
-                    st.subheader("📦 Download Dataset")
-
-                    with open(zip_path, "rb") as f:
-                        st.download_button(
-                            label=f"⬇️ Download Dataset ({zip_size_mb:.1f} MB)",
-                            data=f.read(),
-                            file_name=Path(zip_path).name,
-                            mime="application/zip",
-                            use_container_width=True,
-                            type="primary",
-                        )
-
-                    if session_output.exists():
-                        preview_rgb = session_output / "rgb"
-                        if preview_rgb.exists():
-                            png_files = sorted(preview_rgb.glob("*.png"))
-                            if png_files:
-                                with st.expander("🖼️ Preview", expanded=True):
-                                    preview_cols = st.columns(min(4, len(png_files[:4])))
-                                    for idx, col in enumerate(preview_cols):
-                                        if idx < len(png_files):
-                                            col.image(
-                                                str(png_files[idx]),
-                                                caption=png_files[idx].name,
-                                                use_container_width=True,
-                                            )
-                                    if len(png_files) > 4:
-                                        st.caption(f"... and {len(png_files) - 4} more images")
-
-                    if save_mask:
-                        preview_mask = session_output / "mask"
-                        if preview_mask.exists():
-                            mask_files = sorted(preview_mask.glob("*.png"))
-                            if mask_files:
-                                with st.expander("🏷️ Mask Preview", expanded=True):
-                                    mask_cols = st.columns(min(4, len(mask_files[:4])))
-                                    for idx, col in enumerate(mask_cols):
-                                        if idx < len(mask_files):
-                                            col.image(
-                                                str(mask_files[idx]),
-                                                caption=mask_files[idx].name,
-                                                use_container_width=True,
-                                            )
-
-                st.session_state["last_output_dir"] = str(session_output)
-            else:
-                progress_bar.progress(0.0, text="❌ Generation Failed")
-                st.error(f"Generation failed!\n\n{log_msg}")
-
-        elif uploaded_file is None and start_button:
-            st.warning("Please upload a CAD model file first.")
-
-    st.markdown("---")
-    with st.expander("ℹ️ About", expanded=False):
-        st.markdown("""
-        **Huhb3D Synthetic Data Generator** - Pipeline for generating synthetic training data from CAD models.
-
-        **Workflow:**
-        1. Upload a CAD model (STEP/IGES/STL/OBJ)
-        2. Configure generation parameters
-        3. Click "Start Generation" to run 360° sampling
-        4. Download the ZIP package with RGB images + semantic masks
-
-        **Output Structure:**
-        ```
-        dataset.zip
-        ├── rgb/frame_0001.png ~ frame_NNNN.png
-        ├── mask/mask_0001.png ~ mask_NNNN.png
-        ├── depth/depth_0001.png ~ depth_NNNN.png (if enabled)
-        ├── label_legend.txt
-        ├── camera_poses.json (6DoF camera poses)
-        ├── description.json (if AI description enabled)
-        └── manifest.json
-        ```
-
-        **Semantic Categories:**
-        | ID | Category | Description |
-        |----|----------|-------------|
-        | 0 | FreeSurface | General curved surface |
-        | 1 | HorizontalPlane | Top/bottom face (Y-axis) |
-        | 2 | LateralPlane_X | Side face (X-axis) |
-        | 3 | LateralPlane_Z | Front/back face (Z-axis) |
-        | 4 | NearHorizontal | Near Y-axis surface |
-        | 5 | NearLateral_X | Near X-axis surface |
-        | 6 | NearLateral_Z | Near Z-axis surface |
-        | 7 | Degenerate | Degenerate triangle |
-        | 8 | ConvexFeature_Bolt | Convex protrusion (bolt/boss) |
-        | 9 | ConcaveFeature_Hole | Concave depression (hole) |
-        | 10 | Flange | Flange feature |
-        | 11 | Boss | Boss/stud feature |
-        """)
-
-
-if __name__ == "__main__":
-    main()
+"""
+app.py - Huhb3D Synthetic Data Generator (HuggingFace Space Demo)
+=================================================================
+Demo-only version for HuggingFace Spaces.
+Shows pre-generated sample data: RGB, Mask, Depth, 6DoF, BOP format.
+Full data generation requires local deployment with C++ engine compiled.
+"""
+
+import streamlit as st
+import json
+import os
+from pathlib import Path
+from PIL import Image
+import base64
+from io import BytesIO
+
+DEMO_DIR = Path(__file__).parent / "demo_data"
+
+GITHUB_URL = "https://github.com/AIminminAI/Huhb3D-Viewer"
+GITEE_URL = "https://gitee.com/aiminminai/Huhb3D-Viewer"
+
+
+def image_to_base64(img_path):
+    if not img_path.exists():
+        return None
+    with open(img_path, "rb") as f:
+        return base64.b64encode(f.read()).decode()
+
+
+def load_demo_data():
+    data = {"rgb": [], "mask": [], "depth": [], "manifest": None, "scene_camera": None, "scene_gt": None, "gt_6dof": None, "label_legend": None, "camera_poses": None}
+
+    rgb_dir = DEMO_DIR / "rgb"
+    mask_dir = DEMO_DIR / "mask"
+    depth_dir = DEMO_DIR / "depth"
+
+    if rgb_dir.exists():
+        data["rgb"] = sorted(rgb_dir.glob("*.png"))
+    if mask_dir.exists():
+        data["mask"] = sorted(mask_dir.glob("*.png"))
+    if depth_dir.exists():
+        data["depth"] = sorted(depth_dir.glob("*.png"))
+
+    manifest_path = DEMO_DIR / "manifest.json"
+    if manifest_path.exists():
+        data["manifest"] = json.loads(manifest_path.read_text(encoding="utf-8"))
+
+    scene_camera_path = DEMO_DIR / "scene_camera.json"
+    if scene_camera_path.exists():
+        data["scene_camera"] = json.loads(scene_camera_path.read_text(encoding="utf-8"))
+
+    scene_gt_path = DEMO_DIR / "scene_gt.json"
+    if scene_gt_path.exists():
+        data["scene_gt"] = json.loads(scene_gt_path.read_text(encoding="utf-8"))
+
+    gt_6dof_path = DEMO_DIR / "gt_6dof.json"
+    if gt_6dof_path.exists():
+        data["gt_6dof"] = json.loads(gt_6dof_path.read_text(encoding="utf-8"))
+
+    label_legend_path = DEMO_DIR / "label_legend.txt"
+    if label_legend_path.exists():
+        data["label_legend"] = label_legend_path.read_text(encoding="utf-8")
+
+    camera_poses_path = DEMO_DIR / "camera_poses.json"
+    if camera_poses_path.exists():
+        data["camera_poses"] = json.loads(camera_poses_path.read_text(encoding="utf-8"))
+
+    return data
+
+
+def main():
+    st.set_page_config(
+        page_title="Huhb3D Synthetic Data Generator - Demo",
+        page_icon="🤖",
+        layout="wide",
+    )
+
+    st.title("🤖 Huhb3D Synthetic Data Generator")
+    st.markdown("**面向机器人视觉训练的合成数据生成器** — Demo 展示")
+
+    st.warning(
+        "⚠️ **在线 Demo 模式** — 当前展示预生成的样例数据。"
+        "要使用自己的 CAD 模型生成数据，请在本地部署并编译 C++ 渲染引擎。\n\n"
+        f"📦 GitHub: [{GITHUB_URL}]({GITHUB_URL}) | "
+        f"📦 Gitee: [{GITEE_URL}]({GITEE_URL})"
+    )
+
+    demo = load_demo_data()
+
+    if demo["manifest"]:
+        st.markdown("---")
+        st.subheader("📊 Demo 数据概览")
+        m = demo["manifest"]
+        col1, col2, col3, col4 = st.columns(4)
+        with col1:
+            st.metric("RGB 图像", m.get("rgb_count", len(demo["rgb"])))
+        with col2:
+            st.metric("语义 Mask", m.get("mask_count", len(demo["mask"])))
+        with col3:
+            st.metric("深度图", m.get("depth_count", len(demo["depth"])))
+        with col4:
+            has_6dof = "✅" if demo["gt_6dof"] else "❌"
+            st.metric("6DoF 位姿", has_6dof)
+
+    st.markdown("---")
+    st.subheader("🖼️ 渲染效果展示")
+
+    num_images = min(len(demo["rgb"]), len(demo["mask"]))
+    if num_images == 0:
+        st.info("暂无 Demo 数据。请在本地运行完整版。")
+        return
+
+    selected_idx = st.slider("选择视角", 0, num_images - 1, 0, key="view_slider")
+
+    col_rgb, col_mask, col_depth = st.columns(3)
+
+    with col_rgb:
+        st.markdown("**📷 RGB 渲染图**")
+        if selected_idx < len(demo["rgb"]):
+            st.image(str(demo["rgb"][selected_idx]), use_container_width=True)
+
+    with col_mask:
+        st.markdown("**🏷️ 语义分割 Mask**")
+        if selected_idx < len(demo["mask"]):
+            st.image(str(demo["mask"][selected_idx]), use_container_width=True)
+
+    with col_depth:
+        st.markdown("**📏 深度图**")
+        if selected_idx < len(demo["depth"]):
+            st.image(str(demo["depth"][selected_idx]), use_container_width=True)
+        else:
+            st.info("深度图需在本地编译 C++ 引擎后生成")
+
+    if demo["label_legend"]:
+        with st.expander("🏷️ 语义标签分类", expanded=False):
+            st.text(demo["label_legend"])
+
+    st.markdown("---")
+    st.subheader("📐 6DoF & BOP 格式数据")
+
+    tab_6dof, tab_scene_cam, tab_scene_gt, tab_poses = st.tabs(
+        ["🎯 6DoF Ground Truth", "📷 scene_camera.json", "📋 scene_gt.json", "🎥 camera_poses.json"]
+    )
+
+    with tab_6dof:
+        if demo["gt_6dof"]:
+            st.json(demo["gt_6dof"])
+        else:
+            st.info("6DoF 数据需在本地编译 C++ 引擎后生成")
+
+    with tab_scene_cam:
+        if demo["scene_camera"]:
+            st.json(demo["scene_camera"])
+        else:
+            st.info("BOP scene_camera 数据需在本地生成")
+
+    with tab_scene_gt:
+        if demo["scene_gt"]:
+            st.json(demo["scene_gt"])
+        else:
+            st.info("BOP scene_gt 数据需在本地生成")
+
+    with tab_poses:
+        if demo["camera_poses"]:
+            st.json(demo["camera_poses"])
+        else:
+            st.info("camera_poses 数据需在本地生成")
+
+    st.markdown("---")
+    st.subheader("📋 完整数据格式说明")
+
+    st.markdown("""
+每次生成输出一个 ZIP 包，结构如下：
+
+```
+run_<timestamp>/
+├── rgb/                    # RGB 渲染图 (PNG)
+│   ├── view_000.png
+│   ├── view_001.png
+│   └── ...
+├── mask/                   # 语义分割 Mask (PNG)
+│   ├── view_000_mask.png
+│   └── ...
+├── depth/                  # 深度图 (PNG可视化 + RAW float)
+│   ├── view_000_depth.png
+│   └── view_000_depth.raw
+├── camera_poses.json       # 6DoF 相机位姿
+├── scene_camera.json       # BOP 格式相机内参
+├── scene_gt.json           # BOP 格式 6DoF 位姿
+├── gt_6dof.json            # 完整 6DoF + 四元数
+├── label_legend.txt        # 语义标签对照表
+└── manifest.json           # 数据集元信息
+```
+""")
+
+    st.markdown("---")
+    st.subheader("🔧 本地部署（解锁完整功能）")
+
+    st.markdown(f"""
+**要使用自己的 CAD 模型生成数据，请按以下步骤本地部署：**
+
+1. 克隆仓库：`git clone {GITHUB_URL}`
+2. 安装 Python 依赖：`pip install streamlit Pillow numpy opencv-python`
+3. 编译 C++ 渲染引擎：需要 Visual Studio 2019+ (Windows) 或 GCC (Linux)
+4. 启动 Web UI：`streamlit run app.py --server.port 8501`
+
+**Docker 部署（最省心）：**
+```bash
+docker build -t huhb3d-synthetic .
+docker run -p 7860:7860 huhb3d-synthetic
+```
+
+详见 [GitHub README]({GITHUB_URL}) 获取完整部署指南。
+""")
+
+    st.markdown("---")
+    st.caption("Huhb3D Synthetic Data Generator v2.0 | AGPL-3.0 License")
+
+
+if __name__ == "__main__":
+    main()

demo_data/camera_poses.json

@@ -0,0 +1,38 @@
+[
+  {
+    "frame_id": 0,
+    "position": [0.0, 0.0, 5.0],
+    "rotation_euler": [0.0, 0.0, 0.0],
+    "fov_degrees": 45.0,
+    "view_matrix": [
+      [1.0, 0.0, 0.0, 0.0],
+      [0.0, 1.0, 0.0, 0.0],
+      [0.0, 0.0, 1.0, -5.0],
+      [0.0, 0.0, 0.0, 1.0]
+    ],
+    "projection_matrix": [
+      [2.414, 0.0, 0.0, 0.0],
+      [0.0, 2.414, 0.0, 0.0],
+      [0.0, 0.0, -1.002, -0.200],
+      [0.0, 0.0, -1.0, 0.0]
+    ]
+  },
+  {
+    "frame_id": 1,
+    "position": [1.18, 0.0, 4.86],
+    "rotation_euler": [0.0, -13.6, 0.0],
+    "fov_degrees": 45.0,
+    "view_matrix": [
+      [0.972, 0.0, 0.236, -0.0],
+      [0.0, 1.0, 0.0, 0.0],
+      [-0.236, 0.0, 0.972, -5.0],
+      [0.0, 0.0, 0.0, 1.0]
+    ],
+    "projection_matrix": [
+      [2.414, 0.0, 0.0, 0.0],
+      [0.0, 2.414, 0.0, 0.0],
+      [0.0, 0.0, -1.002, -0.200],
+      [0.0, 0.0, -1.0, 0.0]
+    ]
+  }
+]

demo_data/gt_6dof.json

@@ -0,0 +1,56 @@
+[
+  {
+    "frame_id": 0,
+    "position": [0.0, 0.0, 5.0],
+    "rotation_euler": [0.0, 0.0, 0.0],
+    "rotation_quaternion": [1.0, 0.0, 0.0, 0.0],
+    "fov_degrees": 45.0,
+    "view_matrix": [
+      [1.0, 0.0, 0.0, 0.0],
+      [0.0, 1.0, 0.0, 0.0],
+      [0.0, 0.0, 1.0, -5.0],
+      [0.0, 0.0, 0.0, 1.0]
+    ],
+    "projection_matrix": [
+      [2.414, 0.0, 0.0, 0.0],
+      [0.0, 2.414, 0.0, 0.0],
+      [0.0, 0.0, -1.002, -0.200],
+      [0.0, 0.0, -1.0, 0.0]
+    ],
+    "object_poses": [
+      {
+        "obj_id": 1,
+        "rotation_matrix": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]],
+        "translation": [0.0, 0.0, 0.0],
+        "quaternion": [1.0, 0.0, 0.0, 0.0]
+      }
+    ]
+  },
+  {
+    "frame_id": 1,
+    "position": [1.18, 0.0, 4.86],
+    "rotation_euler": [0.0, -13.6, 0.0],
+    "rotation_quaternion": [0.9929, 0.0, -0.1188, 0.0],
+    "fov_degrees": 45.0,
+    "view_matrix": [
+      [0.972, 0.0, 0.236, -0.0],
+      [0.0, 1.0, 0.0, 0.0],
+      [-0.236, 0.0, 0.972, -5.0],
+      [0.0, 0.0, 0.0, 1.0]
+    ],
+    "projection_matrix": [
+      [2.414, 0.0, 0.0, 0.0],
+      [0.0, 2.414, 0.0, 0.0],
+      [0.0, 0.0, -1.002, -0.200],
+      [0.0, 0.0, -1.0, 0.0]
+    ],
+    "object_poses": [
+      {
+        "obj_id": 1,
+        "rotation_matrix": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]],
+        "translation": [0.0, 0.0, 0.0],
+        "quaternion": [1.0, 0.0, 0.0, 0.0]
+      }
+    ]
+  }
+]

demo_data/label_legend.txt

@@ -0,0 +1,13 @@
+# Semantic Label Color Legend
+# Category -> (R, G, B) in 0-255 range
+
+0 FreeSurface 127 127 127
+1 HorizontalPlane 0 0 255
+2 LateralPlane_X 0 255 0
+3 LateralPlane_Z 255 0 0
+4 NearHorizontal 255 255 0
+5 NearLateral_X 255 0 255
+6 NearLateral_Z 0 255 255
+7 Degenerate 255 127 0
+8 Reserved1 127 0 255
+9 Reserved2 0 127 255

demo_data/manifest.json

@@ -0,0 +1,15 @@
+{
+  "version": "2.0",
+  "generator": "Huhb3D-SyntheticDataPipeline",
+  "rgb_count": 6,
+  "mask_count": 6,
+  "depth_count": 0,
+  "instance_mask_count": 0,
+  "has_legend": true,
+  "has_camera_poses": true,
+  "has_bop_format": true,
+  "has_6dof_gt": true,
+  "has_topology_labels": false,
+  "has_ai_description": false,
+  "note": "Demo data contains 6 sample frames. Full generation requires C++ engine compilation."
+}

demo_data/scene_camera.json

@@ -0,0 +1,16 @@
+{
+  "1": {
+    "cam_K": [1234.0, 0.0, 400.0, 0.0, 1234.0, 300.0, 0.0, 0.0, 1.0],
+    "depth_scale": 1.0,
+    "model_unit": "mm",
+    "cam_R_w2c": [1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0],
+    "cam_t_w2c": [0.0, 0.0, 5000.0]
+  },
+  "2": {
+    "cam_K": [1234.0, 0.0, 400.0, 0.0, 1234.0, 300.0, 0.0, 0.0, 1.0],
+    "depth_scale": 1.0,
+    "model_unit": "mm",
+    "cam_R_w2c": [0.972, 0.0, 0.236, 0.0, 1.0, 0.0, -0.236, 0.0, 0.972],
+    "cam_t_w2c": [0.0, 0.0, 5000.0]
+  }
+}

demo_data/scene_gt.json

@@ -0,0 +1,8 @@
+{
+  "1": [
+    {"obj_id": 1, "cam_R_m2c": [1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0], "cam_t_m2c": [0.0, 0.0, 0.0]}
+  ],
+  "2": [
+    {"obj_id": 1, "cam_R_m2c": [0.972, 0.0, 0.236, 0.0, 1.0, 0.0, -0.236, 0.0, 0.972], "cam_t_m2c": [0.0, 0.0, 0.0]}
+  ]
+}

requirements.txt

@@ -1,4 +1,2 @@
-streamlit>=1.28.0
-requests>=2.28.0
-Pillow>=9.0.0
-numpy>=1.21.0
+streamlit>=1.28.0
+Pillow>=9.0.0