yolo26 seg merge depth detect

Discussion
, , ,
1

I recently combined YOLO26’s segmentation detection with depth detection, using the same backbone network for both and splitting it into two different task heads. A single inference outputs both segmentation and depth information, with the depth calculation referencing the segmentation info. Here’s the GitHub link, feel free to check it out: GitHub - koroandiven/yolo26-seg-depth: yolo26-seg with depth · GitHub

1 Like
2

Thanks for sharing this community experiment. The shared backbone with a frozen segmentation path and mask-guided depth decoder is an interesting multitask extension.

It would be helpful to add the exact Ultralytics version or commit used, downloadable weights, and latency/accuracy comparisons against separate yolo26-seg and yolo26-depth inference. That would make the results in the depth_merge_seg branch easier for others to reproduce and evaluate.

3

the weight is opensource too

4

Thanks for clarifying—you’re right. I confirmed that the repository includes open-source best.pt and last.pt files in the exp16 weights directory. Adding a direct link to the recommended checkpoint in the README would make it easier for users to find and reproduce your results.