A collection of my personal helper tools and scripts for working with images, files and folders in CV tasks.
Split folders with files (e.g. images) into train, validation and test (dataset) folders.
see: split-folders
default: train 80%, val 10%, test 10%
Merge all images from a folder and its subfolders into a single folder. Recombine split images in subfolder structures
Duplicate an image by the specified factor.
Delete random images from a folder. Specify a number of images to be left.
Rename all files. see: https://learn.microsoft.com/de-de/windows/powertoys/powerrename
Note: Activate RegEx
(.*).jpeg
name_${padding=4;increment=1;start=1}
Copy your files faster and more securely
For Windows: https://www.codesector.com/teracopy
Resize images to a new target resolution. Works with a folder with subfolders.
Apply blur and grayscale noise to a given input image.
Create a new dataset with X random versions with blur and grayscale noise of each image. Rename the images and duplicate the annotation label files. This method makes it possible to generate a larger dataset from fewer images by generating variations of the original images and also eliminates the need to generate new annotations by copying the old ones. This is a data augmentation method.
Create a plot to show the first X number of images from a folder in a grid plot.
- Computer Vision Annotation Tool (CVAT): Online or self-hosted with Docker
split-folders
ipykernel
matplotlib
numpy
- Refactor