Skip to content

Multi-object Detection System for Plant Stomata Phenotypic Traits, release only, which has been published on IEEE/ACM-TCBB 2021 (CCF-B)

License

Notifications You must be signed in to change notification settings

zjxi/stomata-auto-detector

Repository files navigation

release License python 3.7

tensorflow keras 2.24 opencv-python pyqt5

Downloads platform downloads stars forks

stomata_auto_detector-win64-v1.2.rar

Example images

example_images

Reference Guide for the Stomata Automatic Detector

1. Overview

  • The Stomata Automatic Detector is a high-throughput microscope image multi-object detection and intelligent recognition system for plant stomatal phenotypic traits, and it is developed by Python3 language based on TensorFlow, Keras, PyQt5, OpenCV frameworks.

  • The Stomata Automatic Detector mainly consists of five modules: File, Detect, Settings, Edit, View, Language, Help.

2. Operation

  • Unzip the downloaded compressed package
  • cd stomata-auto-detector-v1.2
  • The file structure of our system is as below
stomata-auto-detector-xxx
  |__stomata-auto-detector-xxx.exe (the executive file)
  |
  |__resource 
  |         |_____icons (the background pic and logo of system)
  |         |_____images (saved images after single image detection)
  |         |_____info-detected (saved info after single image detection)
  |         |_____param-data (record the values)
  |                        |___model_param_data.txt 
  |                        |___unit_conv_data.txt
  |__utils
         |__json__detection_config.json (configuration of detection)
         |__models__detection_model.h5 (the trained model with the best performance)
  • Double click .exe file, then run the system
  • The detail operations are in Demo video

3. Modules

  • File : It includes the "Open" submenu and the "Quit System" option.

    • the "Open" submenu also contains two options, "Open Image" and "Open Dir".

    • When "Open Image" is selected, an image file dialog will pop up, then select the demand single image, the selected image to be detected will appear in the upper-left area of this system.

    • When "Open Dir" is selected, the same file dialog will appear, and then select the demand image directory that needs to be detected in batches.'

  • Detect : It includes two options, "Detect Image" and "Detect Dir".

    • Refer to the operations of "File" module, These two options could detect the chosen image or dir, respectively.
  • Settings : It includes two options: "Unit Conversion" and "Model Parameter".

    • "Unit Conversion": Calculating the ratio of the scale of the microscope image to its pixel length. The default 100 represents 100 $\mu m$ scale, i.e. 20x magnification in our given maize_20ximages is equivalent to the 100 $\mu m$ scale. The default 433 represents the pixel length of the 100 $\mu m$ black line.

    • "Model Parameter": including "Stomata Minimum Confidence" and "NMS-threshold", both parameter could adjust the accuracy of detection model and eliminate overlapped bouding-boxes.

  • Edit : It includes two options: "Reset System" and "Clean Caches".

    • When "Reset System" option is selected, the system will be reset to the initial state, and several parameters will go back to the default values.

    • When "Clean Caches" option is selected, the cache data saved in the corresponding directory after the stomata detection of the system, so as to release the resources occupied by the system.

  • View : It includes the "Image Preview" option.

    • When click this option, a file dialog will pop up, and then select the demand image folder, and the chosn images will be displayed in two-column alignment.
  • Language : It includes the "English" option.

    • This module represents the corresponding language of the current system.
  • Help : It includes two options: "Website" and "About".

    • When click the "Website" option, the system will automatically jump to the corresponding official website.

    • The "About" option records the sytstem information and copyright details.

Demo video

(bilibili video link)

<iframe src="//player.bilibili.com/player.html?aid=461693047&bvid=BV17L411H7SS&cid=370140226&page=1" scrolling="no" border="0" frameborder="no" framespacing="0" allowfullscreen="true" height="600" width="800"> </iframe>

The dataset is from **National Maize Improvement Center of China**.