feat: Benchmark of crnn_mobilenet_v3 small

[charlesmindee]

This PR adds the benchmark results of crnn_mobilenet_v3_small in TF in the docs, and pytorch results are displayed below.
2 new sets are added, resumes and road fines.

FUNSD

Model Evaluation (model= db_resnet50 + crnn_mobilenet_v3_small, dataset=FUNSD)
Text Detection - Recall: 79.99%, Precision: 87.22%, Mean IoU: 71.29%
Text Recognition - Accuracy: 86.58% (unicase: 87.34%)
OCR - Recall: 68.38% (unicase: 68.83%), Precision: 74.56% (unicase: 75.05%), Mean IoU: 71.29%

CORD
Model Evaluation (model= db_resnet50 + crnn_mobilenet_v3_small, dataset=CORD)
Text Detection - Recall: 93.00%, Precision: 91.37%, Mean IoU: 76.68%
Text Recognition - Accuracy: 92.18% (unicase: 92.60%)
OCR - Recall: 84.30% (unicase: 84.67%), Precision: 82.82% (unicase: 83.18%), Mean IoU: 76.68%

FORM US
Model Evaluation (model= db_resnet50 + crnn_mobilenet_v3_small, dataset=OCRDataset)
Text Detection - Recall: 77.55%, Precision: 91.22%, Mean IoU: 70.77%
Text Recognition - Accuracy: 83.83% (unicase: 84.43%)
OCR - Recall: 74.93% (unicase: 75.23%), Precision: 88.14% (unicase: 88.50%), Mean IoU: 70.77%

IDS
Text Detection - Recall: 69.46%, Precision: 74.21%, Mean IoU: 61.29%
Text Recognition - Accuracy: 65.13% (unicase: 67.93%)
OCR - Recall: 45.67% (unicase: 48.57%), Precision: 48.80% (unicase: 51.89%), Mean IoU: 61.29%

RECEIPTS
Model Evaluation (model= db_resnet50 + crnn_mobilenet_v3_small, dataset=OCRDataset)
Text Detection - Recall: 86.04%, Precision: 89.94%, Mean IoU: 76.12%
Text Recognition - Accuracy: 91.07% (unicase: 92.03%)
OCR - Recall: 77.64% (unicase: 78.45%), Precision: 81.16% (unicase: 82.01%), Mean IoU: 76.12%

INVOICES
Model Evaluation (model= db_resnet50 + crnn_mobilenet_v3_small, dataset=OCRDataset)
Text Detection - Recall: 68.28%, Precision: 72.05%, Mean IoU: 62.69%
Text Recognition - Accuracy: 90.24% (unicase: 91.79%)
OCR - Recall: 61.72% (unicase: 62.57%), Precision: 65.13% (unicase: 66.03%), Mean IoU: 62.69%

ROAD FINES
Model Evaluation (model= db_resnet50 + crnn_mobilenet_v3_small, dataset=OCRDataset)
Text Detection - Recall: 92.34%, Precision: 90.72%, Mean IoU: 83.86%
Text Recognition - Accuracy: 93.76% (unicase: 94.66%)
OCR - Recall: 85.39% (unicase: 86.12%), Precision: 83.90% (unicase: 84.62%), Mean IoU: 83.86%

RESUMES
Model Evaluation (model= db_resnet50 + crnn_mobilenet_v3_small, dataset=OCRDataset)
Text Detection - Recall: 91.51%, Precision: 94.38%, Mean IoU: 86.92%
Text Recognition - Accuracy: 90.86% (unicase: 92.19%)
OCR - Recall: 82.99% (unicase: 84.15%), Precision: 85.60% (unicase: 86.79%), Mean IoU: 86.92%

[codecov]

Codecov Report

Merging #523 (96e2442) into main (09998a4) will decrease coverage by 0.02%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##             main     #523      +/-   ##
==========================================
- Coverage   95.38%   95.36%   -0.03%     
==========================================
  Files         109      109              
  Lines        4183     4183              
==========================================
- Hits         3990     3989       -1     
- Misses        193      194       +1     

Flag	Coverage Δ
unittests	95.36% <ø> (-0.03%)	⬇️

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

Impacted Files	Coverage Δ
...dels/detection/differentiable_binarization/base.py	91.19% <0.00%> (-0.63%)	⬇️

---

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 09998a4...96e2442. Read the comment docs.

[fg-mindee]

Thanks for the benchmarks!