Dmitry Burdeiny
Dmitry Burdeiny
Hi, essentially, only batch size == 1 is supported. The reasons for that are: 1) The input images have different resolutions; 2) After the augmentations that change the input sample...
The code in this repo will work with datasets that have structure compatible with the structure of ShanghaiTech. The readme file (the main page of the repo) shows the directory...
Hi, only batch size == 1 is supported. See https://github.com/dmburd/S-DCNet/issues/9
Hi! You are right. Thanks for pointing out the difference. I will try to fix the bug soon.
Fixed: https://github.com/dmburd/S-DCNet/commit/1d0b07ca1bd833cd59cb3c48b70461113db33406
The exact algorithm for density map generation is not present in the original repos. I followed the procedure described in the [MCNN paper (CVPR 2016)](https://www.cv-foundation.org/openaccess/content_cvpr_2016/papers/Zhang_Single-Image_Crowd_Counting_CVPR_2016_paper.pdf). Values of some of the...
Are you planning to reproduce the density map generation used in the official repo?
Hi, No, I have not tried to convert pytorch -> tensorrt. Another possible option is to try pytorch -> ONNX -> tensorrt conversion.
Please checkout the commit a1179c8. Then edit the script `train.py`: find the line `for sample in self.train_loader:` (should be line 220) and insert 5 lines right above that. ``` print(f"train_loader...
I wonder if the issue is specific to your windows environment. Do you have a Linux OS installation / Linux machine at hand (maybe a virtual machine)? Could you install...