[Feature] Add ImageClassificationInferencer.

[mzr1996]

Motivation

See open-mmlab/mmengine#773 for more details about the inferencer design in MMEngine. And this PR is also depend on the inference branch of MMEngine.

In a word, the inferencer is a enhanced inference_model but split the whole process into preprocess, forward, visualize and postprocess four methods.

Modification

Implement ImageClassificationInferencer, refactor inference_model and our demo.

BC-breaking (Optional)

Almost not, but the demo/image_demo.py is modified and we need to also change our documentation later.

Use cases (Optional)

To use the ImageClassificationInferencer

>>> from mmcls import ImageClassificationInferencer
>>> inferencer = ImageClassificationInferencer('resnet50_8xb32_in1k')
>>> results = inferencer('demo/demo.JPEG')
>>> print(results[0]['pred_class'])
sea snake

To inference multiple images by batch on CUDA

>>> from mmcls import ImageClassificationInferencer
>>> inferencer = ImageClassificationInferencer('resnet50_8xb32_in1k', device='cuda')
>>> imgs = ['demo/demo.JPEG'] * 100
>>> results = inferencer(imgs, batch_size=16)
Inference ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 24.5 it/s
>>> print(results[99]['pred_class'])
sea snake

The original demo/image_demo.py also supports model name by now

python demo/image_demo.py demo/demo.JPEG resnet50_8xb32_in1k

Output:

{
  "pred_label": 65,
  "pred_score": 0.6649367809295654,
  "pred_class": "sea snake"
}

Checklist

Before PR:

• Pre-commit or other linting tools are used to fix the potential lint issues.
• Bug fixes are fully covered by unit tests, the case that causes the bug should be added in the unit tests.
• The modification is covered by complete unit tests. If not, please add more unit test to ensure the correctness.
• The documentation has been modified accordingly, like docstring or example tutorials.

After PR:

• If the modification has potential influence on downstream or other related projects, this PR should be tested with those projects, like MMDet or MMSeg.
• CLA has been signed and all committers have signed the CLA in this PR.

[tonysy]

Do we need to add the shape check of the numpy array?

[tonysy]

How about add some usage cases in doc-string?

[tonysy]

As inline comments

[codecov]

Codecov Report

Base: 0.02% // Head: 86.85% // Increases project coverage by +86.82% 🎉

Coverage data is based on head (16d4a6d) compared to base (b8b31e9).
Patch has no changes to coverable lines.

❗ Current head 16d4a6d differs from pull request most recent head d21e071. Consider uploading reports for the commit d21e071 to get more accurate results

Additional details and impacted files

@@             Coverage Diff              @@
##           dev-1.x    #1261       +/-   ##
============================================
+ Coverage     0.02%   86.85%   +86.82%     
============================================
  Files          121      166       +45     
  Lines         8217    13510     +5293     
  Branches      1368     2148      +780     
============================================
+ Hits             2    11734    +11732     
+ Misses        8215     1420     -6795     
- Partials         0      356      +356     

Flag	Coverage Δ
unittests	86.85% <ø> (+86.82%)	⬆️

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

Impacted Files	Coverage Δ
mmcls/datasets/transforms/compose.py
mmcls/evaluation/metrics/multi_task.py	94.59% <0.00%> (ø)
mmcls/models/retrievers/base.py	100.00% <0.00%> (ø)
mmcls/apis/model.py	87.09% <0.00%> (ø)
mmcls/utils/progress.py	66.66% <0.00%> (ø)
mmcls/models/heads/margin_head.py	89.13% <0.00%> (ø)
mmcls/models/tta/score_tta.py	100.00% <0.00%> (ø)
mmcls/models/utils/norm.py	80.00% <0.00%> (ø)
mmcls/models/backbones/swin_transformer_v2.py	89.63% <0.00%> (ø)
mmcls/models/backbones/levit.py	96.06% <0.00%> (ø)
... and 157 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[tonysy]

As inline comments