Loren's picture
Initial commit
9c06ac7
raw
history blame
1.55 kB
Collections:
- Name: RobustScanner
Metadata:
Training Data: OCRDataset
Training Techniques:
- Adam
Epochs: 5
Batch Size: 1024
Training Resources: 16x GeForce GTX 1080 Ti
Architecture:
- ResNet31OCR
- ChannelReductionEncoder
- RobustScannerDecoder
Paper:
URL: https://arxiv.org/pdf/2007.07542.pdf
Title: 'RobustScanner: Dynamically Enhancing Positional Clues for Robust Text Recognition'
README: configs/textrecog/robust_scanner/README.md
Models:
- Name: robustscanner_r31_academic
In Collection: RobustScanner
Config: configs/textrecog/robust_scanner/robustscanner_r31_academic.py
Metadata:
Training Data:
- ICDAR2011
- ICDAR2013
- ICDAR2015
- COCO text
- IIIT5K
- SynthText
- SynthAdd
- Syn90k
Results:
- Task: Text Recognition
Dataset: IIIT5K
Metrics:
word_acc: 95.1
- Task: Text Recognition
Dataset: SVT
Metrics:
word_acc: 89.2
- Task: Text Recognition
Dataset: ICDAR2013
Metrics:
word_acc: 93.1
- Task: Text Recognition
Dataset: ICDAR2015
Metrics:
word_acc: 77.8
- Task: Text Recognition
Dataset: SVTP
Metrics:
word_acc: 80.3
- Task: Text Recognition
Dataset: CT80
Metrics:
word_acc: 90.3
Weights: https://download.openmmlab.com/mmocr/textrecog/robustscanner/robustscanner_r31_academic-5f05874f.pth