mindspore/model_zoo/official/cv/crnn_seq2seq_ocr

• Contents
• • CRNN-Seq2Seq-OCR Description
  • Model Architecture
  • Dataset
  • Environment Requirements
  • Quick Start
  • Script Description
  • • Script and Sample Code
    • Script Parameters
    • • Training Script Parameters
      • Parameters Configuration
    • Dataset Preparation
  • Training Process
  • • Training
    • • Distributed Training
  • Evaluation Process
  • • Evaluation
• Model Description
• • Performance
  • • Evaluation Performance
    • Inference Performance

Contents

• Contents
  • CRNN-Seq2Seq-OCR Description
  • Model Architecture
  • Dataset
    • Dataset Prepare
  • Environment Requirements
  • Quick Start
  • Script Description
    • Script and Sample Code
    • Script Parameters
      • Training Script Parameters
      • Parameters Configuration
    • Dataset Preparation
  • Training Process
    • Training
      • Distributed Training
  • Evaluation Process
    • Evaluation
  • Model Description
    • Performance
      • Training Performance
      • Evaluation Performance

CRNN-Seq2Seq-OCR Description

CRNN-Seq2Seq-OCR is a neural network model for image based sequence recognition tasks, such as scene text recognition and optical character recognition (OCR). Its architecture is a combination of CNN and sequence to sequence model with attention mechanism.

Model Architecture

CRNN-Seq2Seq-OCR applies a vgg structure to extract features from processed images, following with attention-based encoder and decoder layer, finally utilizes NLL to calculate loss. See src/attention_ocr.py for details.

Dataset

For training and evaluation, we use the French Street Name Signs (FSNS) released by Google as the training data, which contains approximately 1 million training images and their corresponding ground truth words.

Environment Requirements

• Hardware（Ascend）
  • Prepare hardware environment with Ascend processor.
• Framework
  • MindSpore
• For more information, please check the resources below：
  • MindSpore Tutorials
  • MindSpore Python API

Quick Start

• After the dataset is prepared, you may start running the training or the evaluation scripts as follows:

  • Running on Ascend

  # distribute training example in Ascend
  $ bash run_distribute_train.sh [RANK_TABLE_FILE] [DATASET_PATH]
  
  # evaluation example in Ascend
  $ bash run_eval_ascend.sh [DATASET_PATH] [CHECKPOINT_PATH]
  
  # standalone training example in Ascend
  $ bash run_standalone_train.sh [DATASET_NAME] [DATASET_PATH] [PLATFORM]
  

  For distributed training, a hccl configuration file with JSON format needs to be created in advance.

  Please follow the instructions in the link below:
  hccl_tools.

Script Description

Script and Sample Code

crnn-seq2seq-ocr
├── README.md                                   # Descriptions about CRNN-Seq2Seq-OCR
├── scripts
│   ├── run_distribute_train.sh                 # Launch distributed training on Ascend(8 pcs)
│   ├── run_eval_ascend.sh                      # Launch Ascend evaluation
│   └── run_standalone_train.sh                 # Launch standalone training on Ascend(1 pcs)
├── src
│   ├── attention_ocr.py                        # CRNN-Seq2Seq-OCR training wrapper
│   ├── cnn.py                                  # VGG network
│   ├── config.py                               # Parameter configuration
│   ├── create_mindrecord_files.py              # Create mindrecord files from images and ground truth
│   ├── dataset.py                              # Data preprocessing for training and evaluation
│   ├── gru.py                                  # GRU cell wrapper
│   ├── logger.py                               # Logger configuration
│   ├── lstm.py                                 # LSTM cell wrapper
│   ├── seq2seq.py                              # CRNN-Seq2Seq-OCR model structure
│   └── utils.py                                # Utility functions for training and data pre-processing
│   ├── weight_init.py                          # weight initialization of LSTM and GRU
└── train.py                                    # Training script
├── eval.py                                     # Evaluation Script

Script Parameters

Training Script Parameters

# distributed training on Ascend
Usage: bash run_distribute_train.sh [RANK_TABLE_FILE] [DATASET_PATH]

# standalone training
Usage: bash run_standalone_train.sh [DATASET_PATH]

Parameters Configuration

Parameters for both training and evaluation can be set in config.py.

Dataset Preparation

• You may refer to "Generate dataset" in Quick Start to automatically generate a dataset, or you may choose to generate a text image dataset by yourself.

Training Process

• Set options in config.py, including learning rate and other network hyperparameters. Click MindSpore dataset preparation tutorial for more information about dataset.

Training

• Run run_standalone_train.sh for non-distributed training of CRNN-Seq2Seq-OCR model, only support Ascend now.

bash run_standalone_train.sh [DATASET_PATH]

Distributed Training

• Run run_distribute_train.sh for distributed training of CRNN-Seq2Seq-OCR model on Ascend.

bash run_distribute_train.sh [RANK_TABLE_FILE] [DATASET_PATH]

Check the train_parallel0/log.txt and you will get outputs as following:

epoch: 20 step: 4080, loss is 1.56112
epoch: 20 step: 4081, loss is 1.6368448
epoch time: 1559886.096 ms, per step time: 382.231 ms

Evaluation Process

Evaluation

• Run run_eval_ascend.sh for evaluation on Ascend.

bash run_eval_ascend.sh [DATASET_PATH] [CHECKPOINT_PATH]

Check the eval/log and you will get outputs as following:

character precision = 0.967522

Annotation precision precision = 0.635204

Model Description

Performance

Evaluation Performance

Parameters	Ascend
Model Version	V1
Resource	Ascend 910; CPU 2.60GHz, 192cores; Memory 755G; OS Euler2.8
uploaded Date	02/11/2021 (month/day/year)
MindSpore Version	1.2.0
Dataset	FSNS
Training Parameters	epoch=20, batch_size=32
Optimizer	SGD
Loss Function	Negative Log Likelihood
Speed	1pc: 355 ms/step; 8pcs: 385 ms/step
Total time	1pc: 64 hours; 8pcs: 9 hours
Parameters (M)	12
Scripts	crnn_seq2seq_ocr script

Inference Performance

Parameters	Ascend
Model Version	V1
Resource	Ascend 910
Uploaded Date	02/11/2021 (month/day/year)
MindSpore Version	1.2.0
Dataset	FSNS
batch_size	32
outputs	Annotation Precision, Character Precision
Accuracy	Annotation Precision=63.52%, Character Precision=96.75%
Model for inference	12M (.ckpt file)