6D Pose Estimation Proposal

FastSAM Segementation

Uses the FastSAM-x Model for image segmentation. Code from FastSAM github repo for inference.

---

Depth Image And Point Cloud Generation From RGB Image

Pipeline

Using the LiheYoung/depth-anything-Large-hf or intel-isl/MiDaS models a depth image is generated from RGB image.

Then Open3d Library is used to filter and create the point cloud.

Open3d is also used for visualization.

WhatsApp.Video.2025-05-26.at.19.17.20_2a789a25.mp4

Point Cloud