ROBOT_IMAGE_PROCESS_OCR_MODEL_TRAIN_v2.0.1

FOT OCR Model Deployment Project

This project involves the deployment of an FOTS model for OCR (Optical Character Recognition) on-site data collected from digital meters. The pipeline includes preprocessing the collected OCR image data, fine-tuning the FOTS model, and evaluating the model performance. This repository includes code for data preprocessing, training, and model evaluation.

Project Steps

1. Data Import: Import the new data to be processed into the field_data directory.
2. Navigate to OCR Code: Enter the code_ocr folder.

3. Run Evaluation: Execute the evaluation script by running:

   ./eval.sh
   
   

Code and Data Location

• Server URL: 192.168.20.250
• Directory: /data/liudan/ocr

Features

• Data Preprocessing: Scripts to preprocess OCR images before feeding into the model.
• Model Training: Code to fine-tune the FOTS model with the new data.
• Model Evaluation: Scripts to assess the model's performance after fine-tuning.

Contact

For any inquiries, please contact the project team.