AI Paper Atlas

Research map / Detection and segmentation / CNN features

Dataset Enhancement with Instance-Level Augmentations

Orest Kupyn, Christian Rupprecht

ECCV 2024 · Detection and segmentation › CNN features · approach: Joint detection and instance segmentation · 17 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
train with instance-level diffusion repainting augmentation (30% per instance)Deformable-DETR · MS-COCO valAP: 39.3 -> 40.5Improved
train with instance-level augmentationYOLOv5m · MS-COCO valAP: 44.1 -> 45.7Improved
train with instance augmentation in low-data regimesYOLOv5m · MS-COCO (10% data)AP: 25.8 -> 27.9Improved
pretrain on our synthetic VOC data then finetune on real VOC, vs real onlyMask2Former · Pascal VOCmIoU: 77.3 -> 78.2Improved
our synthetic data instead of Diffumask synthetic dataMask2Former · Pascal VOC (synthetic only)mIoU: 57.4 -> 72.9Improved
train saliency models on our DUTS expansionU2Net, F3-Net, TRACER-4 · ECSSD, DUTS-TE, DUT-OMRON, HKU-ISF_max, MAE, S_m: 34/36 metrics improvedImproved
replace all people with synthetic people in COCO trainingDeformable-DETR · COCO val (all classes)AP: 39.3 -> 38.3Hurt
replace people with synthetic peopleDeformable-DETR · COCO val (people only)AP: 49.7 -> 45.6Hurt
anonymize people in DUTSTRACER-0 · DUTS-TEF_max: 0.889 -> 0.886No clear effect
add edge and depth ControlNet conditioning to inpainting (ablation)instance augmentation pipeline · DUTS-TEF_max: 0.886 -> 0.892Improved
add edge control on top of depth (ablation)instance augmentation pipeline · DUTS-TEF_max: 0.890 -> 0.892Improved
add prompt engineering (descriptions, colors, lighting) (ablation)instance augmentation pipeline · DUTS-TEF_max: 0.889 -> 0.892Improved

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