preprocess_for_video_annotation

This repository is made for video annotation preprocess that is uploaded in AI_HUB in Korea. My code transform video to images, which can modify fps, extracting event duration and size

Supporting OS

This project can only run in window
Someday, I will make model that can be run in Ubuntu

code install

pip install -r requirements.txt

code implementation

If you want to convert video in AI_HUB, enter the bellow command.

python video2img.py --src [video dir path] --dst [img dir path] --fps [fps] --wh_size [width height] --duration [duration sec] --extenstion [video extension]

Like this

python video2img.py --src D:\data\first_dataset_2020_10\raw_video\fire --dst D:\data\first_dataset_2020_10\refine_video\fire --fps 10 --wh_size 1280 720 --duration 30 --extension 'mp4'

File structure

It is necessary to set video folder like bellow image
It doesn't have to set xml file(If you have no xml, my model automatically ignore xml file)

src_video_folder
 |
 |_________outsidedoor01
 |              |_________1-5
 |              |          |____1.mp4    
 |              |          |____1.xml
 |              |          |____2.mp4
 |              |          |____2.xml
 |              |
 |              |_________1-6
 |              |          |____1.mp4    
 |              |          |____1.xml
 |              |          |____2.mp4
 |              |          |____2.xml
 |              |_________1-7
 |              
 ....
 |_________outsidedoor02

My model automatically sets the structure like bellow image.
Just know my model how form the structure.

dst_video_folder
 |
 |
 |_________1 img foldoer
 |              |_________img_00000.jpg    
 |              |_________img_00001.jpg
 |              |_________img_00002.jpg
 |              ...
 |_________2 img foldoer
 |              |_________img_00000.jpg    
 |              |_________img_00001.jpg
 |              |_________img_00002.jpg
 |              ...
 |_________100 img foldoer