
ZG
Ziang Gan, Qi Zhu, Libao Zhang
· 1 min read
ResearcharXiv cs.CV
Room for Supervision: Training-Free Dataset Distillation for Object Detection
arXiv:2611.20680v1 Announce Type: new
Abstract: Dataset distillation (DD) reduces storage and training costs by compressing large datasets into small collections. Most existing methods focus on image classification, where each image typically has a single category label. In contrast, object detection requires accurate category and bounding box annotations for multiple objects, together with visual content consistent with those annotations. Existing detection distillation pipelines rely on learned models and can alter the spatial relationships between objects and their backgrounds. We introduce GRASP, which packs dense native supervision into limited pixels without network training or inference. GRASP jointly searches over image selections and spatial layouts using convex allocation subproblems that penalize insufficient object resolution, excessive image compression, and padding that source context cannot fill. During composition, each selected image is cropped around all annotated objects and expanded into surrounding context without reducing object scale. The resized regions are then combined into distilled images, preserving the spatial layout within each region and retaining all original class labels and bounding box annotations. We evaluate GRASP on MS COCO and PASCAL VOC with compression ratios ranging from 0.25\% to 2\%. GRASP constructs the 1\% MS COCO dataset in under two minutes. Under a common training protocol using hard labels, GRASP achieves the state of the art with both CNN and ViT backbones. Code is available at https://github.com/tade1s/GRASP
Original source
This story was published by arXiv cs.CV and written by Ziang Gan, Qi Zhu, Libao Zhang. SyncAI.news shows a preview; the complete article is on the publisher's site.
Read the full story on arxiv.org


