{"id":29420737,"url":"https://github.com/parths007/architect3d","last_synced_at":"2026-02-24T01:05:30.213Z","repository":{"id":303103160,"uuid":"946550654","full_name":"ParthS007/Architect3D","owner":"ParthS007","description":null,"archived":false,"fork":false,"pushed_at":"2025-07-11T09:13:52.000Z","size":49433,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"master","last_synced_at":"2025-12-11T08:35:03.253Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"HTML","has_issues":false,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/ParthS007.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2025-03-11T10:13:04.000Z","updated_at":"2025-07-11T09:13:56.000Z","dependencies_parsed_at":"2025-07-05T19:00:25.975Z","dependency_job_id":null,"html_url":"https://github.com/ParthS007/Architect3D","commit_stats":null,"previous_names":["parths007/architect3d"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/ParthS007/Architect3D","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ParthS007%2FArchitect3D","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ParthS007%2FArchitect3D/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ParthS007%2FArchitect3D/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ParthS007%2FArchitect3D/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/ParthS007","download_url":"https://codeload.github.com/ParthS007/Architect3D/tar.gz/refs/heads/master","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ParthS007%2FArchitect3D/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":29765741,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-02-23T21:02:23.375Z","status":"ssl_error","status_checked_at":"2026-02-23T20:58:31.539Z","response_time":90,"last_error":"SSL_connect returned=1 errno=0 peeraddr=140.82.121.6:443 state=error: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2025-07-12T02:16:45.999Z","updated_at":"2026-02-24T01:05:30.207Z","avatar_url":"https://github.com/ParthS007.png","language":"HTML","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Architect3D: Improving the Mask Quality for Open-Vocabulary Pipelines\n\n\u003cdiv align=\"center\"\u003e\n\n![Architect3D](https://img.shields.io/badge/3D%20Vision-Architect3D-blue?style=for-the-badge)\n![Python](https://img.shields.io/badge/Python-3.8+-brightgreen?style=for-the-badge\u0026logo=python)\n![PyTorch](https://img.shields.io/badge/PyTorch-1.12+-orange?style=for-the-badge\u0026logo=pytorch)\n![CUDA](https://img.shields.io/badge/CUDA-11.3+-green?style=for-the-badge\u0026logo=nvidia)\n\n**Advanced 3D instance segmentation specifically designed for architectural scene understanding**\n\n[📊 Baselines](baseline.md) • [📋 Summary](PROJECT.md) • [🎨 Visualization](interactive_tsne_visualization.html) • [📑 Report](docs/Architect3D.pdf)\n\n\u003c/div\u003e\n\n---\n\n## 🎯 Overview\n\n**Architect3D** adapts the state-of-the-art [Mask3D](https://github.com/JonasSchult/Mask3D) model to work with the **ScanNet++** dataset, enabling fine-grained 3D instance segmentation for architectural scenes with **2,753 classes** - a 10x increase from standard datasets.\n\n### ✨ Key Features\n\n- 🏛️ **Architectural Focus**: Specialized for building and indoor architectural scenes\n- 📈 **Massive Scale**: Handles 2,753 fine-grained architectural classes\n- 🔍 **High Resolution**: Optimized for 0.02m voxel precision\n- 🔗 **OpenMask3D Ready**: Prepared for open-vocabulary integration\n- 📊 **Comprehensive Evaluation**: Detailed architectural scene analysis\n\n### 🚨 Project Status\n\n\u003e **Note**: Due to computational constraints (GPU limitations, 200GB storage limit), full evaluation is pending. The model has been successfully adapted and the framework is complete.\n\n---\n\n## 📊 Performance Overview\n\n| Model | Dataset | Classes | AP | AP50 | AP25 | Status |\n|-------|---------|---------|-----|------|------|--------|\n| Mask3D | ScanNet200 | 200 | 26.9 | 36.2 | 41.4 | ✅ Baseline |\n| OpenMask3D | ScanNet200 | 200 | 15.4 | 19.9 | 23.1 | ✅ Baseline |\n| **Architect3D** | **ScanNet++** | **2,753** | *Pending* | *Pending* | *Pending* | 🔄 Ready |\n\n*See [baseline.md](baseline.md) for detailed comparisons*\n\n---\n\n## 🚀 Quick Start\n\n### Prerequisites\n\n```bash\n# System requirements\nCUDA \u003e= 11.3\nPython \u003e= 3.8\nGPU Memory \u003e= 8GB\n```\n\n### Installation\n\n```bash\n# Clone repository\ngit clone [your-repo-url]\ncd Architect3D\n\n# Install dependencies\npip install -r requirements.txt\n\n# For detailed MinkowskiEngine setup, see below ⬇️\n```\n\n### Basic Usage\n\n```bash\n# 1. Preprocess ScanNet++ data\ncd Architect3D/Mask3D/\nsbatch preprocessing.sh\n\n# 2. Run evaluation\nsbatch scannetpp_eval.sh\n\n# 3. Generate visualizations\npython vis.py\n```\n\n---\n\n## 📁 Repository Structure\n\n\u003cdetails\u003e\n\u003csummary\u003e\u003cstrong\u003e📂 Click to expand detailed structure\u003c/strong\u003e\u003c/summary\u003e\n\n```\nArchitect3D/\n├── 📄 README.md                               # This file\n├── 📋 PROJECT_SUMMARY.md                      # Executive summary\n├── 📊 baseline.md                             # Performance baselines\n├── 🎨 vis.py                                  # t-SNE visualization generator\n├── 🌐 interactive_tsne_visualization.html     # Interactive class embeddings\n├── 📑 Architect3D.pdf                         # Comprehensive project report\n│\n├── 🏗️ Architect3D/                            # Core implementation\n│   ├── Mask3D/                               # Adapted Mask3D for ScanNet++\n│   │   ├── 🚀 main_instance_segmentation.py  # Main training/evaluation script\n│   │   ├── ⚙️ conf/                          # Hydra configurations\n│   │   ├── 📊 benchmark/                     # Evaluation framework\n│   │   ├── 🗃️ datasets/                      # Data loaders \u0026 preprocessing\n│   │   ├── 🧠 models/                        # Neural network architectures\n│   │   ├── 🎯 trainer/                       # Training pipeline\n│   │   ├── 💾 saved/final/                   # Model checkpoints\n│   │   └── 📈 jobs/                          # Training logs\n│   └── 📋 requirements.txt\n│\n├── 🔍 openmask3d/                             # OpenMask3D integration\n│   └── openmask3d/                           # Core modules\n│       ├── 🎭 class_agnostic_mask_computation/\n│       ├── 🔮 mask_features_computation/\n│       ├── 📊 evaluation/\n│       └── 👁️ visualization/\n│\n├── 🏠 scannetpp/                              # ScanNet++ dataset\n│   ├── metadata/                             # Class definitions\n│   ├── scannetpp_ply/                        # 3D scenes\n│   └── splits/                               # Train/val/test splits\n│\n└── 📊 eval_results_architectural_classes/     # Evaluation results\n```\n\n\u003c/details\u003e\n\n---\n\n## 🔧 Technical Implementation\n\n### Architecture Adaptations\n\n```mermaid\ngraph TB\n    A[ScanNet++ Dataset\u003cbr/\u003e2,753 classes] --\u003e B[Sparse 3D CNN\u003cbr/\u003eMinkowskiEngine]\n    B --\u003e C[Multi-scale Features]\n    C --\u003e D[Transformer Decoder]\n    D --\u003e E[Enhanced Head\u003cbr/\u003e2,753 outputs]\n    E --\u003e F[Instance Masks]\n    \n    G[RGB Images] --\u003e H[CLIP Features]\n    H --\u003e I[Multi-view Fusion]\n    I --\u003e J[OpenMask3D Pipeline]\n    \n    F --\u003e K[Architectural\u003cbr/\u003ePredictions]\n    J --\u003e K\n    \n    style A fill:#e1f5fe\n    style E fill:#f3e5f5\n    style K fill:#e8f5e8\n```\n\n### Key Modifications\n\n| Component | Original | Architect3D | Improvement |\n|-----------|----------|-------------|-------------|\n| **Classes** | 200 | **2,753** | 🔥 **13.8x scaling** |\n| **Voxel Size** | 0.05m | **0.02m** | 🎯 **2.5x precision** |\n| **Domain** | General | **Architectural** | 🏛️ **Specialized** |\n| **Head Architecture** | Standard | **Scaled** | ⚡ **Optimized** |\n\n---\n\n## 🛠️ Detailed Installation\n\n\u003cdetails\u003e\n\u003csummary\u003e\u003cstrong\u003e🔧 Complete MinkowskiEngine Setup (ETH Cluster)\u003c/strong\u003e\u003c/summary\u003e\n\n```bash\n# STEP 1: Load modules\nmodule load gcc/8.2.0 python_gpu/3.8.5 cuda/11.3.1 cudnn/8.2.1.32\n\n# STEP 2: Create environment\npython -m venv architect3d_env\nsource architect3d_env/bin/activate\n\n# STEP 3: Install PyTorch\npip install torch==1.12.1 torchvision==0.13.1 -f https://download.pytorch.org/whl/cu113/torch_stable.html\n\n# STEP 4: Install dependencies\npip install ninja pytorch-lightning==1.7.2 hydra-core==1.0.5\n\n# STEP 5: Setup MinkowskiEngine\ngit clone https://github.com/NVIDIA/MinkowskiEngine.git\ncd MinkowskiEngine\n# Edit setup.py (uncomment CUDA_HOME configuration)\npython setup.py install\n\n# STEP 6: Install additional packages\npip install -r requirements.txt\n\n# STEP 7: Install CLIP \u0026 SAM\npip install git+https://github.com/openai/CLIP.git --no-deps\npip install git+https://github.com/facebookresearch/segment-anything.git --no-deps\n```\n\n\u003c/details\u003e\n\n---\n\n## 📈 Evaluation \u0026 Results\n\n### Visualization\n\n🎨 **Interactive t-SNE**: Explore 2,753 architectural class embeddings\n- Open `interactive_tsne_visualization.html` in browser\n- Visualize class relationships and clusters\n- Understand architectural taxonomy\n\n### Metrics\n\n📊 **Comprehensive Evaluation**:\n- **AP Metrics**: Standard instance segmentation evaluation\n- **Class Analysis**: Head/Common/Tail performance breakdown\n- **Architectural Focus**: Building-specific evaluation protocols\n\n---\n\n## 🤝 Acknowledgments\n\n### Core Technologies\n- **[Mask3D](https://github.com/JonasSchult/Mask3D)**: Foundation model\n- **[ScanNet++](https://kaldir.vc.in.tum.de/scannetpp/)**: Enhanced dataset\n- **[OpenMask3D](https://openmask3d.github.io/)**: Open-vocabulary framework\n- **[CLIP](https://github.com/openai/CLIP)**: Vision-language features\n\n### Development\nThis project was developed for the **3D Vision course at ETH Zurich**. Special thanks to supervisors for guidance and the unofficial OpenMask3D codebase.\n\n---\n\n## 📚 Resources\n\n| Resource | Description | Link |\n|----------|-------------|------|\n| 📑 **Full Report** | Comprehensive documentation | [PDF](Final_Report_Architect3D.pdf) |\n| 📊 **Baselines** | Performance comparisons | [Markdown](baseline.md) |\n| 📋 **Summary** | Executive overview | [Summary](PROJECT_SUMMARY.md) |\n| 🎨 **Visualization** | Interactive t-SNE | [HTML](interactive_tsne_visualization.html) |\n| ⚙️ **Configs** | Hydra configuration | [Directory](Architect3D/Mask3D/conf/) |\n\n\n---\n\n\u003cdiv align=\"center\"\u003e\n\n**🏗️ Built for advancing 3D architectural scene understanding 🏗️**\n\n[![ETH Zurich](https://img.shields.io/badge/ETH-Zurich-blue?style=flat-square)](https://ethz.ch)\n[![3D Vision](https://img.shields.io/badge/Course-3D%20Vision-green?style=flat-square)](https://www.cvg.ethz.ch)\n\n\u003c/div\u003e\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fparths007%2Farchitect3d","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fparths007%2Farchitect3d","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fparths007%2Farchitect3d/lists"}