small-models-for-glam/bl-crop-tighten-v1
bl-crop-tighten-v1 Training data for crop tightening on the British Library Book Images collection: 7,565 ABBYY picture-block crops (train 6,050 / validation 757 / test 758) with instance boxes and segmentation masks. The splits are book-safe — no book appears in more than one split (4,484 books total). The labels are weak labels, not human annotations: tiiuae/Falcon-Perception-0.6B ran open-vocabulary segmentation over 8,400 stratified crops (embellishments, plates, medium… See the full description on the dataset page: https://huggingface.co/datasets/small-models-for-glam/bl-crop-tighten-v1.
bl-crop-tighten-v1
Training data for crop tightening on the British Library Book Images collection: 7,565 ABBYY picture-block crops (train 6,050 / validation 757 / test 758) with instance boxes and segmentation masks. The splits are book-safe — no book appears in more than one split (4,484 books total).
The labels are weak labels, not human annotations: tiiuae/Falcon-Perception-0.6B ran open-vocabulary segmentation over 8,400 stratified crops (embellishments, plates, medium, with plain covers as negatives) and a filter pass kept 7,565. Treat them accordingly. Each instance carries bbox ([x, y, w, h], pixels), COCO RLE segmentation, and geometry diagnostics (rectangularity, n_components, frac_largest) used by the filters.
The student trained on this data is small-models-for-glam/bl-crop-tighten-rfdetrseg-clip10; its predictions over the full 1M-image corpus are published as the crop_masks config of the source dataset.
from datasets import load_dataset
ds = load_dataset("small-models-for-glam/bl-crop-tighten-v1", split="train")Built by Daniel van Strien; source images digitised by the British Library (CC0).