According to the code, I can see that the pretrained day-time model is simply the pretrained resnet-101, and I could successfully derive the darkening model from that. However, I failed to reproduce the result of the visual place recognition experiment as the one in the paper (MAP 90.4). Could you please share the trained post-adaptation model? Thanks.
Here is what I got from the model zip file:
According to the code, I can see that the pretrained day-time model is simply the pretrained resnet-101, and I could successfully derive the darkening model from that. However, I failed to reproduce the result of the visual place recognition experiment as the one in the paper (MAP 90.4). Could you please share the trained post-adaptation model? Thanks.