About this repository

This repository is fork of http://github.com/shkh/TrainingAssistant.git

NeoTrainingAssistant is application tool for create annotation data of OpenCV, and to crop images.

Requirements

• python3
• pip3
• git

How to install

1. Clone repository:

    % git clone [email protected]:furaibo/NeoTrainingAssistant.git
   

2. Add Jcrop:

    % cd NeoTrainingAssistant
    % git submodule init
    % git submodule update
    % cd static/Jcrop
    % git checkout master
   

3. Install python modules via pip3

    % sudo pip3 install -r freezed.txt
   

How to run

1. Add images to NeoTrainingAssistant/static/img

2. Run server

    % cd NeoTrainingAssistant
    % python3 views.py
   

This command starts the Flask server on port 5000, and please access http://localhost:5000 with web browser.

How to use

Image

About buttons

Button	Explanation
Back	back to previous image
Undo	remove a latest rectangle
Clear	remove all rectangles
Overall	make a rectangle encloses overall of current image
Dismiss	dismiss(not use) an image and go to next image
Skip	add 'negative' flag to current image and go to next image
Next	add 'positive' flag to current image and go to next image

Get data

After all images will be processed, you will get positive.txt, negative.txt and crop images in NeoTrainingAssistant/static/img_dst/[%DATETIME%]. Dismissed images won't be described in these text-files.

Settings

You can change settings with edit of NeoTrainingAssistant/settings.py.

Flag	Default	Explanation
flag_save_crop	True	get cropped image
flag_report_dump	True	get intermidiate json report
report_dump_interval	3	interval of getting intermidiate json report
flag_remove_src	False	remove source images after process
aspect_ratio	0(free)	ratio of rectangle when cropping
limit_upper_width	600	max width of image in window
limit_upper_height	450	max height of image in window
limit_lower_width	400	min width of image in window
limit_lower_height	300	min height of image in window
alert_click_clear	True	get alert when click 'Clear'
alert_click_overall	True	get alert when click 'Overall'
alert_click_skip	False	get alert when click 'Skip'
alert_click_next	False	get alert when click 'Next'