AI Paper Atlas

Research map / Video / Video segmentation and tracking

Learning Video Object Segmentation With Visual Memory

Pavel Tokmakov, Karteek Alahari, Cordelia Schmid

ICCV 2017 · Video › Video segmentation and tracking · approach: Mask propagation for unsupervised video segmentation · 348 citations (OpenAlex)

Read the paper

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 appearance stream (DeepLab fc6 features) (ablation)visual memory model · DAVIS valmean IoU: 43.5 -> 70.1Improved
use DeepLab features instead of raw RGB (ablation)visual memory model · DAVIS valmean IoU: 58.3 -> 70.1Improved
pretrain appearance net on PASCAL VOC segmentation instead of ImageNet only (ablation)visual memory model · DAVIS valmean IoU: 64.1 -> 70.1Improved
add motion stream (ablation)visual memory model · DAVIS valmean IoU: 59.6 -> 70.1Improved
add ConvGRU memory module (ablation)visual memory model · DAVIS valmean IoU: 64.1 -> 70.1Improved
use ConvGRU instead of ConvRNN (ablation)memory module · DAVIS valmean IoU: 68.7 -> 70.1Improved
use ConvGRU instead of ConvLSTM (ablation)memory module · DAVIS valmean IoU: 68.9 -> 70.1No clear effect
use bidirectional instead of unidirectional processing (ablation)ConvGRU · DAVIS valmean IoU: 67.2 -> 70.1Improved
use fc7 instead of fc6 features (ablation)appearance stream · DAVIS valmean IoU: 70.1 -> 69.8No clear effect
use conv5 instead of fc6 features (ablation)appearance stream · DAVIS valmean IoU: 70.1 -> 67.7Hurt
train on real DAVIS instead of synthetic FT3D with LDOF flow (ablation)visual memory model · DAVIS valmean IoU: 59.6 -> 70.1Improved
add CRF post-processingvisual memory model · DAVIS valmean IoU: 70.1 -> 75.9Improved

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