-
Thank you for publishing the great work!
> We use apex for mixed precision training by default.
> \# do not use mmdet version fp16
Why does the code use apex instead of mmdet's fp16?
Are there…
-
Hello,
I have a question about the mIoU in the training process and test process.
I trained the dataset with validate and val result mIoU was 0.7(for example) after last epoch.
However, when…
-
I trained Swin Transformers Cityscapes Instance Segmentation from pretrained Swin Base on ImageNet 22K.
During training, evaluations are very normal.
E.g:
```
2021-06-29 00:57:35,334 - mmdet - INF…
-
I trained my custom data from cascade_mask_rcnn_swin_tiny_patch4_window7_mstrain_480-800_giou_4conv1f_adamw_3x_coco.py.
Because my data only has bbox mark, without mask mask, so I change all with_mas…
-
I changed crop_size in `config\_base_\dataset\cityscapes.py` from **crop_size = (512, 1024)** to **crop_size = (1024, 2048)** and got a `cuda out of memory` error from colab (free).
- This is the t…
-
-
I trained `cityscapes_detection` for 2 epochs already but loss is always around 1.3-1.4
What should I check to ensure loss going down? Thank you!
This is my config:
```
workflow = [('train', 1…
-
PyTorch: 1.7.1+cu101
PyTorch compiling details: PyTorch built with:
- GCC 7.3
- C++ Version: 201402
- Intel(R) Math Kernel Library Version 2020.0.2 Product Build 20200624 for Intel(R) 64 arc…