AI Paper Atlas

Research map / Video / Video segmentation and tracking

SPLIT: Training-Free AI-Generated and Partially Edited Video Detection via Spatial Patch‑Level Incoherence and Temporal Roughness

Jongyeop Hyun, Hyounghun Kim

ECCV 2026 · Video › Video segmentation and tracking · approach: Video restoration with temporal priors · 3 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 patch-level tokens instead of global CLS embeddings for TTR (ablation)TTR · FakeParts OverallFake Recall at FPR 0.1% (%): 54.21 -> 68.29Improved
add LSMI to patch-level TTR (ablation)SPLIT · FakeParts OverallFake Recall at FPR 0.1% (%): +6.16Improved
increase gamma from 1 to 8 in the fusion rule (ablation)SPLIT · FakeParts OverallFake Recall at FPR 0.1% (%)Improved
increase gamma from 8 to 16 (ablation)SPLIT · FakeParts OverallFake Recall at FPR 0.1% (%)No clear effect
use SPLIT patch-level scoring instead of D3 with same XCLIP-B/16 encoderD3 · FakeParts OverallFake Recall at FPR 0.1% (%): 2.28 -> 74.45Improved
use XCLIP-B/16 instead of XCLIP-B/32 encoder (ablation)SPLIT · FakeParts OverallFake Recall at FPR 0.1% (%): 60.23 -> 74.45Improved
use CLIP-B/16 instead of CLIP-B/32 encoder (ablation)SPLIT · FakeParts OverallFake Recall at FPR 0.1% (%): 47.37 -> 71.47Improved
apply Gaussian blur sigma=2 to input videosSPLIT · FakeParts OverallFake Recall at FPR 0.1% (%): 74.45 -> 55.42Hurt
apply JPEG compression q=80SPLIT · FakeParts OverallFake Recall at FPR 0.1% (%): 74.45 -> 64.40Hurt
apply vertical flip to input videosSPLIT · FakeParts OverallFake Recall at FPR 0.1% (%): 74.45 -> 66.92Hurt
apply horizontal flip to input videosSPLIT · FakeParts OverallFake Recall at FPR 0.1% (%): 74.45 -> 75.73No clear effect
replace D3 with SPLIT on ResNet18 encoderD3 · FakeParts OverallFake Recall at FPR 0.1% (%): 1.93 -> 61.23Improved

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