search

alanlukezic / d3s

D3S - Discriminative Single Shot Segmentation Tracker (CVPR 2020)
265 stars 57 forks source link

d3s on davis2016 #23

Open riversci opened 3 years ago

riversci commented 3 years ago

Hi, I think d3s is an excellent work which combined both tracking and segmentation. I like it very much. However, when I tested the pretrained model on davis2016, I got results as follow. image it's not as same as the paper.
image I am confused. Can you give me some advice?

alanlukezic commented 3 years ago

One possible reason for different scores is different hardware. We observed that the performance can differ due to different GPUs or even library versions. The other possible reason is tracker initialization. It should be initialized as: tracker.initialize(img, gt_rect, init_mask=gt_mask) where gt_rect is an 8-element vector of the rectangle corners: [x0, y0, x1, y1, x2, y2, x3, y2].

riversci commented 3 years ago

Thank you for your answer. But How you get the gt_rect from gt_mask? Can you share some details?

laisimiao commented 1 year ago

@alanlukezic could you write a guide doc to help us evaluate D3S on davis such vos datasets?

laisimiao commented 1 year ago

@riversci how you test on davis2016, could you give some guidences?