Upgrade to mmdet 2.20.0

[ai-fast-track]

I upgraded https://github.com/airctic/mmdetection_configs to mmdet 2.20.0

This PR upgrade mmdet_config_path as well as the pip install mmdet==2.20.0

closes #1040

[codecov[bot]]

Codecov Report

Merging #1041 (8fda0dc) into master (04644ac) will decrease coverage by 0.21%. The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #1041      +/-   ##
==========================================
- Coverage   85.93%   85.71%   -0.22%     
==========================================
  Files         278      278              
  Lines        5915     5915              
==========================================
- Hits         5083     5070      -13     
- Misses        832      845      +13     

Flag	Coverage Δ
unittests	85.71% <100.00%> (-0.22%)	:arrow_down:

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
icevision/models/mmdet/common/utils.py	81.81% <ø> (ø)
icevision/models/mmdet/download_configs.py	80.95% <100.00%> (ø)
icevision/models/inference_sahi.py	81.44% <0.00%> (-13.41%)	:arrow_down:

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 04644ac...8fda0dc. Read the comment docs.

[dnth]

Hi everyone, I notice that with mmdet==2.20.0 the VFNet fails to train with the following error. Any mmdet version above 2.17.0 shows the same error.

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-16-35383c9197b5> in <module>()
----> 1 learn.fine_tune(20, 0.00158, freeze_epochs=1)

21 frames
/usr/local/lib/python3.7/dist-packages/mmcv/runner/fp16_utils.py in new_func(*args, **kwargs)
    184                                 'method of nn.Module')
    185             if not (hasattr(args[0], 'fp16_enabled') and args[0].fp16_enabled):
--> 186                 return old_func(*args, **kwargs)
    187             # get the arg spec of the decorated method
    188             args_info = getfullargspec(old_func)

TypeError: loss() missing 1 required positional argument: 'img_metas'

Here is a Colab notebook I ran https://colab.research.google.com/github/airctic/icevision/blob/master/notebooks/getting_started_object_detection.ipynb

[potipot]

I can confirm having the same error

[ai-fast-track]

@dnth @potipot I experienced the same issue. We need to investigate that!