image-dataset-augmentatator

Image augmentation pipepline for conditional GANs base on Augmentator library

Available parameters

• --input_dir path to directory containing the images
• --output_dir path to directory to save new images
• --output_format output image format (JPEG|BMP|PNG|GIF)
• --output_width output image width in pixels
• --output_height output image height in pixels
• --resample_filter filter to use when resampling (BICUBIC|BILINEAR|ANTIALIAS|NEAREST)
• --sample_count amount of new images to generate
• --zoom zoom in randomly
• --distortion distort randomly
• --output_prefix prefix used to lable output images 'prefix_00001.jpeg'

Basic execution

python3 augmentator.py --input_dir 'path/to/your/images' --output_dir 'path/to/your/augmented/images' --zoom --distortion --output_prefix 'relevantprefix' --sample_count 100 

This will do random crops and resize the images to 512px also add random zoom and distortion.