AI Paper Atlas

Research map / Video / Video segmentation and tracking

Synthetic Visual Genome 2: Extracting Large-scale Spatio-Temporal Scene Graphs from Videos

Ziqi Gao, Jieyu Zhang, Wisdom Ikezogwo, Jae Sung Park, Tario You, Daniel Ogbu, Chenhao Zheng, Weikai Huang, Yinuo Yang, Winson Han, Quan Kong, Rajat Saini and 1 more

ECCV 2026 · Video › Video segmentation and tracking · approach: Layered video decomposition and dynamics · 4 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 trajectory-aligned dual-resampler design with Qwen2.5-VL-3B instead of plain fine-tuning on bbox trajectoriesQwen2.5-VL-3B · SVG2_test (lenient)object accuracy: 46.1 -> 79.0Improved
use TraSeR instead of GPT-5video scene graph generation · SVG2_test (lenient)object accuracy: 65.5 -> 79.0Improved
use TraSeR instead of GPT-5 for relationsvideo scene graph generation · PVSG (lenient, IoU 0.5)relation recall: 18.3 -> 16.9Hurt
add temporal-window resampler on top of object-trajectory resampler (ablation)TraSeR · PVSG* (VidOR subset)relation recall: 2.68 -> 11.38Improved
add object-trajectory resampler on top of temporal-window resampler (ablation)TraSeR · PVSG* (VidOR subset)object accuracy: 77.84 -> 78.21No clear effect
add object-trajectory resampler alone (no window resampler) (ablation)TraSeR · PVSG* (VidOR subset)relation recall: 7.15 -> 2.68Hurt
use segmentation instead of bounding-box token arrangement (ablation)TraSeR · PVSG* (VidOR subset)relation recall: 9.39 -> 11.38Improved
train on SVG2 (SA-V) in addition to academic datasets (ablation)TraSeR · SVG2_testobject accuracy: 25.68 -> 78.14Improved
add PVD subset to training data (ablation)TraSeR · SVG2_testattribute recall: 22.72 -> 27.10Improved
train on academic only vs full data for VidOR relations (ablation)TraSeR · VidORrelation recall: 28.66 -> 25.01Hurt
add TraSeR scene graphs to video input for GPT-4.1GPT-4.1 VQA · Perception-Testaccuracy: 66.8 -> 71.4Improved
add TraSeR scene graphs to video input for GPT-4.1GPT-4.1 VQA · AGQA 2.0accuracy: 25.9 -> 26.3Improved

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