ECCV22-P3AFormer-Tracking-Objects-as-Pixel-wise-Distributions icon indicating copy to clipboard operation
ECCV22-P3AFormer-Tracking-Objects-as-Pixel-wise-Distributions copied to clipboard

Published 20 hours ago • verified publisher icon dvlab-research

two question

Open fuyi02 opened this issue 3 years ago • 1 comments

  1. load model, get error: "PytorchStreamReader failed reading zip archive: failed finding central directory" load scripts is sh configs/standard/v100_test_mot17.sh
  2. I can't find code about pixel-wise association, only bytetracker part, no center points association which present in your paper., Is this code complete? image image

fuyi02 avatar Aug 04 '22 09:08 fuyi02

Yes, the same issue is here! The paper introduces the inspired idea from the flow-guided feature aggregation for video object detection which I could not be able to find the usage inside the code.

@dense_tracker.py file the get_flow has no usage. from util.p3aformer.p3aformer_misc import ( get_flow, NestedTensor, gaussian_radius, affine_transform, draw_umich_gaussian, soft_nms_pytorch, )

smart-face avatar Sep 29 '22 07:09 smart-face