🔪 Segmentation
Learning Geometry-Disentangled Representation for Complementary Understanding of 3D Object Point Cloud. (AAAI2021)
PointNet and PointNet++ implemented by pytorch (pure python) and on ModelNet, ShapeNet and S3DIS.
ROS/ROS 2 package for Ultralytics YOLOv8 real-time object detection and segmentation. https://github.com/ultralytics/ultralytics
ROS wrapper for Meta's Segment-Anything model
Grounded SAM: Marrying Grounding DINO with Segment Anything & Stable Diffusion & Recognize Anything - Automatically Detect , Segment and Generate Anything
Track-Anything is a flexible and interactive tool for video object tracking and segmentation, based on Segment Anything, XMem, and E2FGVI.
Grounded Tracking for Streaming Videos
[ECCV 2024] Official implementation of the paper "Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection"
Official Implementation of CVPR24 highlight paper: Matching Anything by Segmenting Anything
[ECCV 2024] Better Call SAL: Towards Learning to Segment Anything in Lidar
Real-Time 3D Semantic Reconstruction from 2D data
This is the official code for MobileSAM project that makes SAM lightweight for mobile applications and beyond!
[ICLR 2025, Oral] EmbodiedSAM: Online Segment Any 3D Thing in Real Time