AI Paper Atlas

Research map / Detection and segmentation / CNN features

Salient Object Subitizing

Jianming Zhang, Shugao Ma, Mehrnoosh Sameki, Stan Sclaroff, Margrit Betke, Zhe Lin, Xiaohui Shen, Brian L. Price, Radomír Měch

IJCV 2017 · Detection and segmentation › CNN features · approach: RGB salient object detection · 87 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
add two-stage pre-fine-tuning on synthetic images (ablation)GoogleNet SOS CNN · SOS test setmean AP: 78.6 -> 80.4Improved
add synthetic images by mixing with real data instead of pre-training (ablation)GoogleNet SOS CNN · SOS test setmean AP: 78.6 -> 77.8No clear effect
fine-tune the whole network instead of only the output layer (ablation)GoogleNet · SOS test setmean AP: 61.3 -> 78.6Improved
synthetic pre-training for 3-object images (ablation)GoogleNet SOS CNN · SOS test setAP for 3 objects: 58.6 -> 64.3Improved
synthetic pre-training for 0-object images (ablation)GoogleNet SOS CNN · SOS test setAP for 0 objects: 93.6 -> 93.5No clear effect
use CNN features instead of SIFT+IVFGoogleNet without fine-tuning · SOS test setmean AP: 50.1 -> 61.3Improved
add synthetic pre-training to AlexNet (ablation)AlexNet SOS CNN · SOS test setmean AP: 70.1 -> 71.6Improved
add synthetic pre-training to VGG16 (ablation)VGG16 SOS CNN · SOS test setmean AP: 77.5 -> 80.2Improved
add synthetic pre-training with only 25% real data (ablation)GoogleNet SOS CNN · SOS test set (25% real data)mean AP: 71.6 -> 76.3Improved
add synthetic pre-training with 50% real data (ablation)GoogleNet SOS CNN · SOS test set (50% real data)mean AP: 75.3 -> 78.2Improved
replace GoogleNet with AlexNet (ablation)SOS CNN · SOS test setmean AP: 78.6 -> 70.1Hurt
cue unconstrained salient object detection with predicted object countUSOD · MSOF-score: 75.7 -> 79.5Improved

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