Image Tools

Batch image process tools mainly designed for dataset preprocessing

Current tools

Global args

• thread The thread pool size for parallel operations

Common

• crop-rectangle Crop a rectangle region of the image
• normalize Normalize image to given range
• map-color Map one RGB color to another in a given PNG image file
• map-background-color Map all the non-valid color in the image to a given color
• split-images Split large images to small pieces for augmentation purposes
• split-images-with-bias Split large images to small pieces for augmentation purposes with bias (Bias is added between each split)
• split-images-with-filter Split large images to small pieces with a filter for enough valid pixels
• class2rgb Map 8 bit grayscale PNG class image to RGB image
• rgb2class Map RGB image to 8 bit grayscale PNG class image
• resize-images Resize all images in a given folder to a given size with a given filter
• rgb2rle Convert RGB semantic segmentation PNG labels to RLE format
• split-dataset Split dataset into train and test sets
• count-classes Count class for 8 bit PNG image & Calc class balance weight
• strip-image-edge Strip image edges
• stich-images Stich the splited images back together
• calc-mean-std Calc the mean and std of a dataset for normalization
• calc-iou Calc the IoU of two images

Yolo

• split-dataset Split dataset into train and test sets Will store result in TXT file
• count-types Count the object number of each type in the dataset
• rgb2yolo Convert RGB labels to YOLO TXT format