[Model] ABCNet Recognizer

[gaotongxiao]

This PR is based on #746 and #747. This recognizer is standalone and is ready to be trained on any text recognition dataset.

Modifications

1. A new NLLLoss module
2. An example toy config

[codecov[bot]]

Codecov Report

Merging #811 (c2e4ecb) into main (fb77352) will decrease coverage by 0.01%. The diff coverage is 84.92%.

@@            Coverage Diff             @@
##             main     #811      +/-   ##
==========================================
- Coverage   84.61%   84.60%   -0.02%     
==========================================
  Files         165      169       +4     
  Lines       10674    10787     +113     
  Branches     1624     1635      +11     
==========================================
+ Hits         9032     9126      +94     
- Misses       1297     1315      +18     
- Partials      345      346       +1     

Flag	Coverage Δ
unittests	84.60% <84.92%> (-0.02%)	:arrow_down:

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

Impacted Files	Coverage Δ
mmocr/models/textrecog/decoders/crnn_decoder.py	75.00% <ø> (ø)
mmocr/models/textrecog/losses/nll_loss.py	29.16% <29.16%> (ø)
mmocr/models/textrecog/layers/rnn_layers.py	95.83% <95.83%> (ø)
...ocr/models/textrecog/decoders/abc_recog_decoder.py	100.00% <100.00%> (ø)
...ocr/models/textrecog/encoders/abc_recog_encoder.py	100.00% <100.00%> (ø)
mmocr/models/textrecog/recognizer/abcnet.py	100.00% <100.00%> (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 fb77352...c2e4ecb. Read the comment docs.