AI Paper Atlas

Research map / Video / Video segmentation and tracking

DeepVS: A Deep Learning Based Video Saliency Prediction Approach

Lai Jiang, Mai Xu, Tie Liu, Minglang Qiao, Zulin Wang

ECCV 2018 · Video › Video segmentation and tracking · approach: Stochastic video prediction · 203 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 auxiliary KL loss on the cross-net mask (ablation)OM-CNN · LEDOVKL divergence: decreases KL by 0.24Improved
combine objectness and motion subnets into OM-CNN (vs objectness subnet alone) (ablation)OM-CNN · LEDOVKL divergence: 0.05 reduction in KL divergenceImproved
combine objectness and motion subnets into OM-CNN (vs motion subnet alone) (ablation)OM-CNN · LEDOVKL divergence: 0.09 KL divergence reductionImproved
add SS-ConvLSTM on top of OM-CNN (ablation)DeepVS · LEDOVKL divergence: reduces KL by 0.09Improved
add center-bias dropout (p=0.75) instead of no CB dropout (ablation)SS-ConvLSTM · LEDOV validationKL divergence: reduce KL divergence by 0.03Improved
lower dropout rates from 0.75 to 0.2 (ablation)SS-ConvLSTM · LEDOV validationKL divergence: KL rises by 0.08Hurt
test with longer clips (24 or 32 frames) than training length 16 (ablation)SS-ConvLSTM · LEDOV validationKL divergence: a bit better at 24 or 32No clear effect
use DeepVS instead of best baselineDeepVS · LEDOV testAUC, NSS, CC, KL: 0.01, 0.51, 0.12 and 0.33 improvementsImproved
transfer DeepVS trained on LEDOV to DIEMDeepVS · DIEMAUC, NSS, CC, KL: 0.05, 0.57, 0.11 and 0.34 improvementsImproved
transfer DeepVS trained on LEDOV to SFUDeepVS · SFUCC, KL: CC 0.55 vs 0.58; KL 0.67 vs GBVS 0.61Mixed

Later papers that compared with this method

2 later accepted papers report DeepVS 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