search

hikvision-research / ProbabilisticTeacher

An official implementation of ICML 2022 paper "Learning Domain Adaptive Object Detection with Probabilistic Teacher"."
Apache License 2.0
70 stars 10 forks source link

About the results #2

Closed chenhaocs closed 1 year ago

chenhaocs commented 1 year ago

Hi, thanks for sharing the codes. Will you update the links of weights and log files in the "Main Results" ?

merlinarer commented 1 year ago

Hi, thanks for sharing the codes. Will you update the links of weights and log files in the "Main Results" ?

Yes, c2f is already there, and others may take a while. Several emails sent to me says the results are reproduced with this code, and you can run it following the instruction.

chenhaocs commented 1 year ago

https://github.com/hikvision-research/ProbabilisticTeacher/blob/ea34bca288f4609c1882439540cfda5de4bf2ce5/pt/modeling/roi_heads/fast_rcnn.py#L250

Another question, there's some difference about the formular between the code and the paper, could you explain the reason ?

merlinarer commented 1 year ago

https://github.com/hikvision-research/ProbabilisticTeacher/blob/ea34bca288f4609c1882439540cfda5de4bf2ce5/pt/modeling/roi_heads/fast_rcnn.py#L250

Another question, there's some difference about the formular between the code and the paper, could you explain the reason ?

Here is the KL-divergence, equal to the CE during optimization.

kinredon commented 1 year ago

When to release the weights. and log files for other adaptation scenarios? @merlinarer