Improved 3D Reconstruction Algorithm & Point Cloud Processing (SfM-MVS / NeRF / 3DGS)
Budget: $30 – $250 USD
I'm looking for an expert to improve an existing 3D reconstruction algorithm and enhance point cloud processing.
• Computer Vision（计算机视觉）
• 3D Modelling（三维建模）
• Algorithm（算法）
• Deep Learning（深度学习）
• Image Processing（图像处理）
• Machine Learning (ML)
• Python（SfM/NeRF/3DGS 基于 Python）
• C++ Programming（ SfM-MVS 库如 COLMAP、OpenMVS C++）
• OpenCV（图像特征提取、SfM）
• Photogrammetry
• Point Cloud
• 3D Rendering
• Tensorflow / PyTorch
• Neural Networks
Computer Vision, 3D Modelling, Algorithm, Deep Learning, Image Processing, Machine Learning (ML), Python, OpenCV, Photogrammetry, Point Cloud
Key Requirements:
- Focus on SfM, MVS, or NeRF
- Aim for higher accuracy, faster processing, or better noise handling
- Input data can be images, video, or LiDAR
Ideal Skills and Experience:
- Strong background in computer vision and 3D reconstruction
- Proficiency in handling and optimizing large datasets
- Experience with point cloud processing tools and frameworks
- Ability to provide measurable improvements in algorithms
Please share your approach, relevant experience, and any similar projects completed.
Improved 3D Reconstruction Algorithm & Point Cloud Processing (SfM-MVS / NeRF / 3DGS)
Looking for experts in 3D reconstruction and point cloud processing (SfM-MVS / NeRF / 3DGS). Task: reconstruct transmission towers & powerlines from drone images, generate measurable point clouds/meshes, extract anchor points, and improve over baseline methods.
I need an improved 3D reconstruction algorithm for power transmission towers and wires, based on UAV photographs (images will be provided).
Project Requirements:
1. Measurement Target
• Reconstruct towers and wires in 3D.
• Extract 3D coordinates of specific anchor points and calculate their horizontal distance and height difference.
2. Data Conditions
• Input: UAV photographs (with GPS only, no RTK).
• No absolute scale — a known reference length (e.g. insulator string length, cross arm size) will be used to compute the scale factor.
3. Technical Requirements
• Must provide an improved algorithm (not just vanilla SfM).
• Output: measurable point cloud (.ply/.las) or mesh (.obj/.ply).
• Possible approaches:
– Enhanced SfM-MVS (line features, semantic segmentation, cable curve fitting).
– Advanced methods such as 3D Gaussian Splatting (3DGS) or NeRF.
• Results should be better than baseline SfM/NeRF/3DGS, with quantitative evaluation.
4. Deliverables
• Source code with documentation.
• Scaled point cloud / mesh.
• A measurement tool/script: select two points → output 3D coordinates, horizontal distance, and height difference.
• A short report comparing results with baseline methods.
This is an academic-level research project (not commercial 3D modeling). Looking for freelancers with strong experience in computer vision, 3D reconstruction, SfM/NeRF/3DGS, or point cloud processing, ideally with prior work on powerlines or thin structure reconstruction.
Data file --->> Data file.docx
• Computer Vision（计算机视觉）
• 3D Modelling（三维建模）
• Algorithm（算法）
• Deep Learning（深度学习）
• Image Processing（图像处理）
• Machine Learning (ML)
• Python（SfM/NeRF/3DGS 基于 Python）
• C++ Programming（ SfM-MVS 库如 COLMAP、OpenMVS C++）
• OpenCV（图像特征提取、SfM）
• Photogrammetry
• Point Cloud
• 3D Rendering
• Tensorflow / PyTorch
• Neural Networks
Computer Vision, 3D Modelling, Algorithm, Deep Learning, Image Processing, Machine Learning (ML), Python, OpenCV, Photogrammetry, Point Cloud
Key Requirements:
- Focus on SfM, MVS, or NeRF
- Aim for higher accuracy, faster processing, or better noise handling
- Input data can be images, video, or LiDAR
Ideal Skills and Experience:
- Strong background in computer vision and 3D reconstruction
- Proficiency in handling and optimizing large datasets
- Experience with point cloud processing tools and frameworks
- Ability to provide measurable improvements in algorithms
Please share your approach, relevant experience, and any similar projects completed.
Improved 3D Reconstruction Algorithm & Point Cloud Processing (SfM-MVS / NeRF / 3DGS)
Looking for experts in 3D reconstruction and point cloud processing (SfM-MVS / NeRF / 3DGS). Task: reconstruct transmission towers & powerlines from drone images, generate measurable point clouds/meshes, extract anchor points, and improve over baseline methods.
I need an improved 3D reconstruction algorithm for power transmission towers and wires, based on UAV photographs (images will be provided).
Project Requirements:
1. Measurement Target
• Reconstruct towers and wires in 3D.
• Extract 3D coordinates of specific anchor points and calculate their horizontal distance and height difference.
2. Data Conditions
• Input: UAV photographs (with GPS only, no RTK).
• No absolute scale — a known reference length (e.g. insulator string length, cross arm size) will be used to compute the scale factor.
3. Technical Requirements
• Must provide an improved algorithm (not just vanilla SfM).
• Output: measurable point cloud (.ply/.las) or mesh (.obj/.ply).
• Possible approaches:
– Enhanced SfM-MVS (line features, semantic segmentation, cable curve fitting).
– Advanced methods such as 3D Gaussian Splatting (3DGS) or NeRF.
• Results should be better than baseline SfM/NeRF/3DGS, with quantitative evaluation.
4. Deliverables
• Source code with documentation.
• Scaled point cloud / mesh.
• A measurement tool/script: select two points → output 3D coordinates, horizontal distance, and height difference.
• A short report comparing results with baseline methods.
This is an academic-level research project (not commercial 3D modeling). Looking for freelancers with strong experience in computer vision, 3D reconstruction, SfM/NeRF/3DGS, or point cloud processing, ideally with prior work on powerlines or thin structure reconstruction.
Data file --->> Data file.docx