Junyu Gao

如果采用我们的Matlab生成密度图，误差很小，没必要去归一化。对于性能提升有限。 如果采用了某些公开的python代码生成密度图，可能会与原有人数有较大差别。或许做一个归一化会有更好的结果。

About SHHA's training, you'd better adopt our provided parameters. For testing, you can preprocess the data using this [code](https://github.com/gjy3035/C-3-Framework/blob/python3.x/datasets/QNRF/preapre_QNRF.m).

we will add the following description: In this code, the validation is directly on the test set. Strictly speaking, it is evaluated on the val set (randomly selected from the...

1080Ti x4, 12G Memory. Training on single GPU works well using a smaller batch size.

https://github.com/pytorch/vision/blob/master/torchvision/models/resnet.py

For SE Cycle GAN, we add a loss in Cycle GAN during the reconstruction process. In practice, we add a .py file and modify the model file. Considering the minor...

Because of many deadlines in March and April, we don't have enough time to check the entire SE Cycle GAN code and release it. After April, we will try our...

Other settings are followed by [pix2pix's](https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix/tree/pytorch0.3.1) (pytorch 0.3.1) default configuration (the paper also mentions it). To be specific, - lambda is 10.0; - size setting: - pre process: resize GCC...

maybe compare the setting.py and config.py in the results reports folder.

I upload the code that achieves MAE of 7.7, you can run it. Note that it is a python2 version. [code.zip](https://github.com/gjy3035/C-3-Framework/files/3958910/code.zip)