Skip to content

Latest commit

 

History

History
52 lines (32 loc) · 2.66 KB

README.md

File metadata and controls

52 lines (32 loc) · 2.66 KB
Raw

handwritten-simplified-chinese-recognition-0001

Use Case and High-Level Description

This is a network for handwritten simplified Chinese text recognition scenario. The model consists of three parts, namely the residual convolutional neural network (CNN) as the feature extractor, and then a flatten operation followed by a fully connected layer as the classifier for final prediction. The network is able to recognize simplified Chinese text consisting of characters in the SCUT-EPT dataset.

Example

-> 的人不一了是他有为在责新中任自之我们

Specification

Metric Value
GFlops 134.513
MParams 17.270
Accuracy on SCUT-EPT test subset (excluding images wider than 2000px after resized to height 96px with aspect ratio) 75.31%
Source framework PyTorch*

Accuracy Values

This model adopts label error rate as the metric for accuracy.

Inputs

Grayscale image, name - actual_input, shape - 1, 1, 96, 2000, format is B, C, H, W, where:

  • B - batch size
  • C - number of channels
  • H - image height
  • W - image width

NOTE: the source image should be resized to specific height (such as 96) while keeping aspect ratio, and the width after resizing should be no larger than 2000 and then the width should be right-bottom padded to 2000 with edge values.

Outputs

Name - output, shape - 125, 1, 4059, format is W, B, L, where:

  • W - output sequence length
  • B - batch size
  • L - confidence distribution across the supported symbols in SCUT-EPT.

The network output can be decoded by CTC Greedy Decoder.

Demo usage

The model can be used in the following demos provided by the Open Model Zoo to show its capabilities:

Legal Information

[*] Other names and brands may be claimed as the property of others.