Drew outlines for the hairpiece. This is a step that I had forgotten to do after remaking the dataset after version 42.
Pre-processed the dataset using a script that preserves the original image's aspect ratio by padding the canvas with transparent pixels rather than cropping.
These changes should improve the odds of accurately recreating the hairpiece during inference.
Drew outlines for the hairpiece. This is a step that I had forgotten to do after remaking the dataset after version 42.
Pre-processed the dataset using a script that preserves the original image's aspect ratio by padding the canvas with transparent pixels rather than cropping.
These changes should improve the odds of accurately recreating the hairpiece during inference.