https://github.com/oil3/ml-py-tools

Collection of scripts for dataset management, bbox convert, mass rename, mass image resize.. Coco single class/category download, folder sorting... Interactive scripts, positional arguments, minimum imports, nothing external to download
https://github.com/oil3/ml-py-tools

annotation-converter bbox coreml labels resize-images script yolo

Last synced: about 2 months ago
JSON representation

Collection of scripts for dataset management, bbox convert, mass rename, mass image resize.. Coco single class/category download, folder sorting... Interactive scripts, positional arguments, minimum imports, nothing external to download

• Host: GitHub
• URL: https://github.com/oil3/ml-py-tools
• Owner: Oil3
• Created: 2024-02-07T01:40:33.000Z (11 months ago)
• Default Branch: main
• Last Pushed: 2024-10-18T04:55:36.000Z (3 months ago)
• Last Synced: 2024-10-20T18:53:43.501Z (3 months ago)
• Topics: annotation-converter, bbox, coreml, labels, resize-images, script, yolo
• Language: Python
• Homepage:
• Size: 54.7 KB
• Stars: 2
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
**yobb_to_yolo_bbox** script to convert orientated bbox labels into normal yolo bbox. Interactive, multithreaded, checks for errors, keeps going, if given, optionally, the images folder, it checks that images have their pairs. Backup is created and then labels overwritten.  

**mass_rename.py** -  script to give same random name to each label/image pairs `python3 script.py labels_path images_path`  

  

**resize.py** - interactive script to resize images inside a folder using PIL, choice of interpolation nearest, bilinear, bicubic, lanczos. "all" option to compare. 

**download_coco_categories.py** - downloads specific classes from the coco dataset, multi-threaded, also creates yolo-format label files for each image.

![image](https://github.com/user-attachments/assets/33877648-ee9f-423f-9fa8-2103ba61dc1a)

Interactive as in prompts you what and where with then a progress bar, multi-threaded as in as fast as possible, also runnable for exemple, like `py script.py 0` for class 0 in a newfolder in current directoy, or like `py script.py 1,33,56,57,70 download_path` with specific directory. 

  

**Structure.py** - recreates folder structure while choosing how many files are copied from the original folders.

I use to create validation datesets from training datasets.

  

**Json_to_Folders.py** - sorts files into folders based on selected attributes values of a JSON file.

It prompts for the number of files to process in each folder, with the option to process all files.

  

**videocrop_path_top_right_bot_left.py** - Crops a video file  by removing specified amounts of pixels from its sides (top, right, bottom, left).

`py script.py video.mp4 100 0 200 20` would create a video_cropped.mp4 with 100 pixels removed from the top, 0 pixels removed from the right side, 200 pixels removed from the bottom, and 20 pixels removed from the left side.

  

**nospaces.swift** - removes python-breaking characters of files within chosen folder, and replaces spaces by underscores.  

`swift nospaces.swift` asks for folder if not given as argument. Logs changes (old name - new name) in a txt file within folder and just had to be a Swift script.