search

LoraLinH / Semi-supervised-Crowd-Counting-via-Density-Agency

Official Implement of ACM MM 2022 paper 'Semi supervised Crowd Counting via Density Agency'
20 stars 5 forks source link

What dose label_count mean? #4

Closed mast1ren closed 1 year ago

mast1ren commented 1 year ago

label_count = torch.tensor( [0.00016, 0.0048202634789049625, 0.01209819596260786, 0.02164922095835209, 0.03357841819524765, 0.04810526967048645, 0.06570728123188019, 0.08683456480503082, 0.11207923293113708, 0.1422334909439087, 0.17838051915168762, 0.22167329490184784, 0.2732916474342346, 0.33556100726127625, 0.41080838441848755, 0.5030269622802734, 0.6174761652946472, 0.762194037437439, 0.9506691694259644, 1.2056223154067993, 1.5706151723861694, 2.138580322265625, 3.233219861984253, 7.914860725402832]) in file regression_trainer.py, line 85. Is it a pre-count data of the dataset used? How can I get it if I want to train another dataset?

LoraLinH commented 1 year ago

The setting is following UEPNet (https://github.com/TencentYoutuResearch/CrowdCounting-UEPNet)

mast1ren commented 1 year ago

I see. Thank u!