AI Paper Atlas

Research map / Video / Video segmentation and tracking

A New Large Scale Dynamic Texture Dataset with Application to ConvNet Understanding

Isma Hadji, Richard P. Wildes

ECCV 2018 · Video › Video segmentation and tracking · approach: Learned optical flow estimation · 28 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
use MSOE-filtered input instead of optical flow for the motion stream (ablation)Flow stream · DTDB-Dynamicsaccuracy: 72.6 -> 80.1Improved
use MSOE-filtered input instead of optical flow for the motion stream (ablation)Flow stream · DTDB-Appearanceaccuracy: 64.8 -> 72.2Improved
fuse RGB appearance stream with the MSOE stream (ablation)MSOE stream · DTDB-Dynamicsaccuracy: 80.1 -> 84.0Improved
fuse RGB appearance stream with the MSOE stream (ablation)MSOE stream · DTDB-Appearanceaccuracy: 72.2 -> 80.0Improved
pretrain on DTDB-dynamics instead of UCF-101MSOE-two-stream · YUVL-1 feature transferaccuracy: 80.8 -> 90.7Improved
pretrain on DTDB-appearance instead of UCF-101MSOE-two-stream · YUP++ feature transferaccuracy: 74.6 -> 89.6Improved
pretrain on DTDB-appearance instead of DTDB-dynamicsMSOE-two-stream · YUVL-1 feature transferaccuracy: 90.7 -> 89.8Hurt
pretrain on DTDB-appearance instead of DTDB-dynamicsMSOE-two-stream · YUP++ feature transferaccuracy: 87.7 -> 89.6Improved
finetune pretrained MSOE-two-stream on DTDB-appearanceMSOE-two-stream · YUP++ entireaccuracy: 89.6 -> 91.8Improved
finetune pretrained C3D on DTDB-dynamics instead of using DTDB-trained-from-scratch featuresC3D · YUVL-1accuracy: 83.3 -> 89.1Improved
camera motion present (moving vs static camera)MSOE-two-stream trained on DTDB-appearance · YUP++ feature transferaccuracy: 94.7 -> 83.2Hurt
use hand-crafted SOE-Net instead of learned MSOE-two-streamMSOE-two-stream · DTDB-Dynamicsaccuracy: 84.0 -> 86.8Improved

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