AI Paper Atlas

Research map / Detection and segmentation / CNN features

FLSL: Feature-level Self-supervised Learning

Qing Su, Anton Netchaev, Hai Li, Shihao Ji

NeurIPS 2023 · Detection and segmentation › CNN features · approach: Joint detection and instance segmentation · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
increase gamma from 1 to 5 (ablation)FLSL · bbox-aligned k-NN (ViT-S)k-NN top-1: 66.6 -> 72.4Improved
remove Sinkhorn normalization (gamma=1) (ablation)FLSL · bbox-aligned k-NN (ViT-S)k-NN top-1: 71.2 -> 66.6Hurt
vary upsilon from 0.01 to 0.1 (ablation)FLSL · bbox-aligned k-NN (ViT-S)k-NN top-1: 68.7 -> 71.2 -> 65.1Mixed
increase number of centroids K from 1024 to 4096 (ablation)FLSL · bbox-aligned k-NNk-NN top-1: 68.1 -> 72.4Improved
increase K from 4096 to 16384 (ablation)FLSL · bbox-aligned k-NNk-NN top-1: 72.4 -> 72.1No clear effect
use FLSL instead of DINO pretrainingViTDet-B/16 · COCO Mask R-CNNAP bbox: 48.9 -> 53.1Improved
use FLSL instead of MAE pretrainingViTDet-B/16 · COCO Mask R-CNNAP mask: 45.5 -> 47.0Improved
use FLSL instead of ADCLRViT-S/16 Mask R-CNN · COCOAP bbox: 44.3 -> 44.9Improved
use FLSL instead of DINO+SelfPatchViT-S/16 Mask R-CNN · COCOAP bbox: 42.1 -> 44.9Improved
reduce patch size from 16 to 8 for FLSLViT-S Mask R-CNN · COCOAP bbox: 44.9 -> 46.5Improved
use FLSL instead of DINO pretrainingViT-S/8 Faster R-CNN · UAVDTAP: 51.1 -> 55.2Improved
use FLSL instead of DINO pretrainingViT-B/16 Faster R-CNN · UAVDTAP: 49.1 -> 53.5Improved

Later papers that compared with this method

1 later accepted paper reports FLSL as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy