[Feature] support torchserve

[nijkah]

Motivation

closes #38
Support to deploy to torchserve

Modification

add tools and sample Dockerfile

Use cases (Optional)

Checklist

1. Pre-commit or other linting tools are used to fix the potential lint issues.
2. The modification is covered by complete unit tests. If not, please add more unit test to ensure the correctness.
3. The documentation has been modified accordingly, like docstring or example tutorials.

• test tools/deployment/mmrotate2serve.py with configs/rotated_faster_rcnn/rotated_faster_rcnn_r50_fpn_1x_dota_le90.py
• test Dockerfile and torchserve
• add a documentation

[zytx121]

Hi @nijkah
Good job! Thanks for your PR.

[codecov-commenter]

Codecov Report

Merging #160 (141a62c) into dev (3a2dfbf) will increase coverage by 0.42%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##              dev     #160      +/-   ##
==========================================
+ Coverage   28.88%   29.30%   +0.42%     
==========================================
  Files         109      113       +4     
  Lines        7074     7298     +224     
  Branches     1064     1110      +46     
==========================================
+ Hits         2043     2139      +96     
- Misses       4961     5083     +122     
- Partials       70       76       +6     

Flag	Coverage Δ
unittests	29.28% <ø> (+0.41%)	⬆️

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

Impacted Files	Coverage Δ
mmrotate/apis/train.py	0.00% <0.00%> (ø)
mmrotate/utils/__init__.py	100.00% <0.00%> (ø)
mmrotate/models/dense_heads/__init__.py	100.00% <0.00%> (ø)
mmrotate/core/bbox/assigners/__init__.py	100.00% <0.00%> (ø)
mmrotate/core/bbox/iou_calculators/__init__.py	100.00% <0.00%> (ø)
mmrotate/models/dense_heads/rotated_atss_head.py	12.65% <0.00%> (ø)
mmrotate/utils/setup_env.py	92.00% <0.00%> (ø)
mmrotate/core/bbox/assigners/atss_obb_assigner.py	17.46% <0.00%> (ø)
mmrotate/utils/compat_config.py	78.33% <0.00%> (ø)

---

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 3a2dfbf...141a62c. Read the comment docs.

[nijkah]

Note that

1. I added brief document for torchserve in English. But I cannot speak Chinese. Any help would be appreciated.
2. There is a version issue in torchserve. See this issue. It should be fixed in mmdetection too. There are three options. I followed the second one.

• 1. Upgrade the version of torch>=1.8.1
• 2. Limit the version of torchserve==0.2.0
• 3. Wait for the latest torchserve to fix the issue.

3. There is a segmentation postprocessing part in tools/deployment/mmrotate_handler.py. Should I delete this?

[liuyanyi]

LGTM

[yangxue0827]

Docstring is required, like-wise for preprocess, inference and postprocess.

[yangxue0827]

LGTM

[zytx121]

LGTM