FaBR Precise157

FaBR Precise157 FaBR Precise157 is a 157-image hard-case bird-mask benchmark for post-mask correction and fine-structure semantic segmentation. This dataset is the formal benchmark used in the FaBR paper experiments. It is not the full 5,000+ image source annotation project. Code for the FaBR mainline pipeline is available here: ShallM/fabr-mainline The benchmark is meant to be small, readable, and easy to inspect. It contains the hard cases used for paper-facing evaluation… See the full description on the dataset page:

Source: Hugging Face Hub (MitchelMi/fabr-precise157). Metadata imported from the dataset’s Hub tags.