RT-DeepLoc localizes temporal deepfakes via reconstruction errors from an MAE trained solely on authentic data plus an asymmetric intra-video contrastive loss, reporting SOTA results on LAV-DF under video-level supervision.
ActionFormer: Localizing Moments of Actions with Transformers
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Mining Forgery Traces from Reconstruction Error: A Weakly Supervised Framework for Multimodal Deepfake Temporal Localization
RT-DeepLoc localizes temporal deepfakes via reconstruction errors from an MAE trained solely on authentic data plus an asymmetric intra-video contrastive loss, reporting SOTA results on LAV-DF under video-level supervision.