Autofocus-RNN

中文

Introduction

Implementation of 《Autofocus of whole slide imaging based on convolution and recurrent neural networks》.

Requirement

• PyTorch 1.1.0
• PIL
• OpenCV2

Dataset

Baida Pan link: https://pan.baidu.com/s/1w8P_1iloZrqw-XeeuTUooQ Extraction code: nn2u

Model Parameters

Baida Pan link: https://pan.baidu.com/s/1bZfugCtaq83EkUlpwp1QEA Extraction code: bqf8

Usage Guide

Dataset Processing

After downloading and extracting the dataset, utilize the tools in the dataset/tools directory to convert images into the structure required for training.

1. Construct the focus_measures tool in the dataset/tools/focus_measures directory, which relies on OpenCV2 and CMake as the build tool.

2. Use the Python scripts located in the dataset/tools directory to generate JSON files recording dataset information. The calc_focus_measures.py script employs the tool created in Step 1 to compute focus measures and saves the data in JSON files for ease of use during model training.

Training/Evaluating the Model

Configure config.py, primarily setting the dataset path and specifying the training, validation, and testing datasets.

Execute train.py or evaluate.py to train or test model.

Note

This project is no longer maintained. After several years, the author finds it challenging to recall the implementation details of the code. 🐶