AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Common Sense Reasoning for Deep Fake Detection

Yue Zhang, Ben Colman, Xiao Guo, Ali Shahriyari, Gaurav Bharaj

ECCV 2024 · Vision-language and multimodal › Multimodal LLMs · approach: Interleaved image-text generation with MLLMs

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 text contrastive loss to BLIP fine-tuning on DD-VQA (ablation)BLIP · DD-VQA testdetection accuracy: 0.8168 -> 0.8365Improved
add image contrastive loss to BLIP fine-tuning on DD-VQA (ablation)BLIP · DD-VQA testdetection accuracy: 0.8168 -> 0.8487Improved
add both text and image contrastive losses (ablation)BLIP · DD-VQA testdetection F1: 0.8642 -> 0.9007Improved
add recall consideration: contrastive losses reduce recall (ablation)BLIP · DD-VQA testdetection recall: 0.9596 -> 0.9341Hurt
integrate DD-VQA representations (BLIP-TI) into XceptionNetXceptionNet · FF++ (c40) intra-testingAUC: 89.30 -> 92.24Improved
integrate DD-VQA representations into HiFi-NetHiFi-Net · WildDeepfake cross-testingAUC: 65.22 -> 70.03Improved
integrate DD-VQA representations into SBIsSBIs · Celeb-DF cross-testingAUC: 93.18 -> 93.98Improved
integrate DD-VQA representations into RECCERECCE · WildDeepfake cross-testingAUC: 64.31 -> 69.46Improved
use plain pretrained BLIP weights in RECCE instead of no BLIP (ablation)RECCE · WildDeepfake cross-testingAUC: 64.31 -> 67.24Improved
use plain pretrained BLIP weights in RECCE (ablation)RECCE · FF++ (c40) intra-testingACC: 91.03 -> 89.22Hurt
add textual explanations to answers (ablation)BLIP on DD-VQA · DD-VQA testdetection accuracy: 0.7978 -> 0.8168Improved
add textual explanations to answers, cross-tested (ablation)BLIP · SeqDeepfake cross-testingdetection accuracy: 0.5216 -> 0.5648Improved

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