search

liruilong940607 / Pose2Seg

Code for the paper "Pose2Seg: Detection Free Human Instance Segmentation" @ CVPR2019.
http://www.liruilong.cn/projects/pose2seg/index.html
MIT License
532 stars 136 forks source link

The instance segmentation performance rely heavily on keypoints performance #19

Open zdwong opened 5 years ago

zdwong commented 5 years ago

Pose2Seg model provided by author was used to generate instance segmentation on Crowdhuman dataset with the input of keypoints generated by AlphaPose model. According to the visualization result, the instance segmentation is not so good compared to mask rcnn model because the performance of instance segmentation of Pose2Seg depended heavily on keypoints performance.

The keypoints visualization: 273275,8192f000acfb8e7b

The instance segmentation visualization: 273275,8192f000acfb8e7b

I was amazed when I saw the visualization of instance segmentation in COCO or COHuman dataset. But when using another model to generate keypoints, the result is not good.

Tetsujinfr commented 5 years ago

do you have the Mask RCNN inferred version of this image? just to eyeball both? cause indeed the pixel segmentation you have with pose2seg is not super impressive regarding details, but it does manage occlusion pretty well I would say, no?

Xuan-YE commented 4 years ago

Pose2Seg model provided by author was used to generate instance segmentation on Crowdhuman dataset with the input of keypoints generated by AlphaPose model. According to the visualization result, the instance segmentation is not so good compared to mask rcnn model because the performance of instance segmentation of Pose2Seg depended heavily on keypoints performance.

The keypoints visualization: 273275,8192f000acfb8e7b

The instance segmentation visualization: 273275,8192f000acfb8e7b

I was amazed when I saw the visualization of instance segmentation in COCO or COHuman dataset. But when using another model to generate keypoints, the result is not good.

Hi! I'm interested in the work you have done. Could you show me some details about how to run it on my own set of images? I'm very grateful to you!