FCOS
FCOS copied to clipboard
FCOS: Fully Convolutional One-Stage Object Detection (ICCV'19)
Fixes [this issue](https://github.com/tianzhi0549/FCOS/issues/357) AT_CHECK is [deprecated since torch 1.5](https://github.com/pytorch/pytorch/issues/36581), so i added this code for compatibility.
Newer python versions [do not have variable `torch._six.PY3`](https://github.com/pytorch/pytorch/blob/master/torch/_six.py#L27) [[Issue](https://github.com/tianzhi0549/FCOS/issues/358)] which is also a module not meant to be imported in a project. PY3 only checks that python version is 3,...
AttributeError: module 'torch._six' has no attribute 'PY3' In file [fcos_core/utils/imports.py](https://github.com/tianzhi0549/FCOS/blob/master/fcos_core/utils/imports.py#L4) there's an if checking if torch._six.PY3. Latest torch version doesn't have PY3 but [PY37](https://github.com/pytorch/pytorch/blob/master/torch/_six.py#L27)
The following lines are the errors: running build running build_py running egg_info writing fcos.egg-info/PKG-INFO writing dependency_links to fcos.egg-info/dependency_links.txt writing requirements to fcos.egg-info/requires.txt writing top-level names to fcos.egg-info/top_level.txt reading manifest file...
```python C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\bin\HostX86\x64\link.exe /nologo /INCREMENTAL:NO /LTCG /DLL /MANIFEST:EMBED,ID=2 /MANIFESTUAC:NO /LIBPATH:C:\Users\Administrator\anaconda3\envs\yol o\lib\site-packages\torch\lib "/LIBPATH:C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1\lib/x64" /LIBPATH:C:\Users\Administrator\anaconda3\envs\yolo\libs /LIBPATH:C:\Users\Administrator\anaconda3\envs\yolo\PCbuild\amd64 "/LIBP ATH:C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\ATLMFC\lib\x64" "/LIBPATH:C:\Program Files (x86)\Microsoft...
E:\workProject\FCOS-master>python setup.py build develop --no-deps running build running build_py running egg_info writing fcos.egg-info\PKG-INFO writing dependency_links to fcos.egg-info\dependency_links.txt writing requirements to fcos.egg-info\requires.txt writing top-level names to fcos.egg-info\top_level.txt reading manifest file 'fcos.egg-info\SOURCES.txt'...
In the code the implementation of the Giou loss is different than the original code: w_intersect = torch.min(pred_left, target_left) + torch.min(pred_right, target_right) g_w_intersect = torch.max(pred_left, target_left) + torch.max( pred_right, target_right)...
Does this version of code support win10?Or just support linux?
I have converted other datasets to VOC format, can I train it? if so.can you explain in detail how to modify which part of the code. Thank you very much!
Hi, I had trained FCOS model on my own dataset and I evaluate the model using the modified coco_eval.py The performance is very low but the predicted boxes are weird....