The detection and localization of highly realistic deepfake audio-visual content are challenging even for the most advanced state-of-the-art methods. The comprehensive benchmark of the proposed dataset utilizing state-of-the-art deepfake detection and localization methods indicates a significant drop in performance compared to previous datasets.
1 PAPER • NO BENCHMARKS YET
…Moreover, eleven state-of-the-art algorithms are evaluated on the benchmark using two evaluation metrics, with detailed analysis provided for the evaluation results.
6 PAPERS • NO BENCHMARKS YET
…Using this, we created two unique NR-VQA models: (a) a local-to-global region-based NR VQA architecture (called PVQ) that learns to predict global video quality and achieves state-of-the-art performance
13 PAPERS • 1 BENCHMARK
…Video recordings were postprocessed using the state-of-the-art pose estimator BodyTrack (similar to OpenPose, and included in NVIDIA Maxine-AR-SDK) to provide a sequence of 3D joint positions for each
0 PAPER • NO BENCHMARKS YET
…Additionally, the small sizes make them only marginally usable for state-of-the-art learning systems, such as deep learning.
17 PAPERS • 1 BENCHMARK
…We hope that the SoccerTrack dataset will help advance the state of the art in multi-object tracking, especially in team sports.