Create dataset from videos

You want to create a dataset from videos?

Then, may be, this code will help you. Cheers!

Install the Requirements

• pip install -r requierments.txt

Hotkeys

Key	Purpose
P	Play/Pause video
Z/X	Rewind +/- 60 frames
C/V	Rewind +/- 3 frames
B/N	Rewind +/- 600 frames
U/I	Rewind +/- 10000 frames
. or ,	Next/Previous video
S	Save Frame
D	Delete last saved frame

Usage

1. In screenshot_from_video.py edit all necessary variables, look for:

• data_path - may be path/to/folder with videos or path/to/video.ext
• file_extension - file/files extension
• skip_to - if video is too long u can skip via timecode

2. Run screenshot_from_video.py.

Or use argparse

3. Execute in bash

python screenshot_from_video.py --videos "/path/to/videos/" --ext ".mp4" --skip "3:14:15" 

or if you want only one video to procces

python screenshot_from_video.py --videos "/path/to/video.mp4" --ext ".mp4" --skip "3:14:15"

4. The script will automatically create folder dataset in "some_folder", where your photos will be.
   • base_folder
     • your_video(s)
     • dataset
       • your_images

Bonus Scripts

• yolo_scripts/select_yolo_labels.py - Select specific yolo labels to generate new labels folder. Then you can use new folder to split data for training.

• yolo_scripts/split_dataset.py - Splitting dataset and preparing to train YoloV5.

• yolo_scripts/coco2yolo.py - Converts COCO annotation json to yolo labels.

• yolo_scripts/yolo2coco.py - Converts yolo label back to readible json with backwards bbox transformation.

• yolo_scripts/coco_json_parser.py - Useful tool to prepare your data after markup. For more info check the file

• mkv_to_mp4.py - Convert mkv files to mp4

• opencv_build.txt - Helpful file if you want to build opencv-python from source. For example: if you want to use h264 (*avc1) codec for videos.

For more info use --help flag or check the files.