Awesome_3D_Vision

📒Motivation

• Tracking the cutting-edge 3D Reconstruction and Generation technology, including 3D Gaussian Splatting, SLAM, Neural Radiance Fields.
• Analyzing the State-of-the-art papers from top conferences, including but not limited to SIGGRAPH, CVPR, ICCV, ECCV, ICLR, ICML and NIPS.
• Replicating classical paper results with SOTA performance.
• Integrating the state-of-the-art algorithms and deploy it directly on the edge devices.

📙TODO

• [2024.6.17]🔥 Based on Infusion, fix a bug of it, supplement the code and open majority of source codes.
• Integrating SAGA, LangSplat into our system, for 3D semantic segmentation / open-vocabulary segmentation tasks.
• Integrating LGM, LucidDreamer into our system, for 3D object generation task.
• Integrating Mip-Splatting into our system, for eliminating multiple artifacts and achieving alias-free renderings.

Survey and Papers

• 🔥 A Survey on 3D Gaussian Splatting ⭐️⭐️⭐
• 🔥DreamGaussian: Generative Gaussian Splatting for Efficient 3D Content Creation⭐️⭐️

Experiments Visualization

Based on Infusion, 3D inpainting task is supported in our system. Original scene:

After removing target object:

Initial scene by lifting points to fill unmasked region:

Finetuned scene:

Deploying Tutorial

3D inpainting