AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

Scene Parsing via Integrated Classification Model and Variance-Based Regularization

Hengcan Shi, Hongliang Li, Qingbo Wu, Zichen Song

CVPR 2019 · Detection and segmentation › Open-set and few-shot detection · approach: Object-level representation learning · 14 citations (OpenAlex)

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 an integrated refinement classifier to PSPNet (ablation)PSPNet · NYU Depth v2mIoU: 46.9 -> 50.3Improved
add variance-based regularization to the baseline (ablation)PSPNet · NYU Depth v2mIoU: 46.9 -> 47.8Improved
combine integrated classification with variance-based regularization (ablation)PSPNet · NYU Depth v2mIoU: 46.9 -> 50.7Improved
add DSN auxiliary classifiers (ablation)PSPNet · NYU Depth v2mIoU: 46.9 -> 46.5Hurt
add more layers and parameters to the baseline (ablation)PSPNet · NYU Depth v2pAcc.: 73.6 -> 73.3Hurt
use multiple binary rather than multinomial general classifiers (ablation)integrated classification model · NYU Depth v2mIoU: 49.4 -> 50.7Improved
use the general-classification probability map as refinement input (ablation)refinement classification network · NYU Depth v2mIoU: 46.9 -> 50.1Improved
also provide image features and the original image to refinement (ablation)refinement classification network · NYU Depth v2mIoU: 50.1 -> 50.7Improved
set the variance-regularization weight to 0.2 (ablation)integrated classification model · NYU Depth v2mIoU: λvbr=0: 50.3; λvbr=0.2: 50.7; λvbr=2: 49.7Mixed
use a three-layer refinement network with 1024-dimensional hidden layers (ablation)integrated classification model · NYU Depth v2mIoU: 50.7Mixed
use the proposed integrated classifier and variance regularizationscene parsing models · PASCAL-ContextmIoU: 51.7 -> 52.6Improved
use the proposed integrated classifier and variance regularizationscene parsing models · SUN-RGBDmIoU: 47.1 -> 50.6Improved

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