Papers › Localizing Moments in Long Video Via Multimodal Guidance

Localizing Moments in Long Video Via Multimodal Guidance

26 Feb 2023ICCV 2023 1arXiv:2302.13372archive 2025-07-28

Wayner Barrios, Mattia Soldan, Alberto Mario Ceballos-Arroyo, Fabian Caba Heilbron, Bernard Ghanem

The recent introduction of the large-scale, long-form MAD and Ego4D datasets has enabled researchers to investigate the performance of current state-of-the-art methods for video grounding in the long-form setup, with interesting findings: current grounding methods alone fail at tackling this challenging task and setup due to their inability to process long video sequences. In this paper, we propose a method for improving the performance of natural language grounding in long videos by identifying and pruning out non-describable windows. We design a guided grounding framework consisting of a Guidance Model and a base grounding model. The Guidance Model emphasizes describable windows, while the base grounding model analyzes short temporal windows to determine which segments accurately match a given language query. We offer two designs for the Guidance Model: Query-Agnostic and Query-Dependent, which balance efficiency and accuracy. Experiments demonstrate that our proposed method outperforms state-of-the-art models by 4.1% in MAD and 4.52% in Ego4D (NLQ), respectively. Code, data and MAD's audio features necessary to reproduce our experiments are available at: https://github.com/waybarrios/guidance-based-video-grounding.

PaperPDFConference PDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

waybarrios/guidance-based-video-grounding officialmentioned in papermentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Natural Language Moment RetrievalNatural Language Visual GroundingVideo GroundingVideo Understanding

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Natural Language Moment Retrieval MAD Zero-Shot CLIP + Guidance Model R@1,IoU=0.1 9.3 #4 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD Zero-Shot CLIP + Guidance Model R@1,IoU=0.3 4.65 #4 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD Zero-Shot CLIP + Guidance Model R@1,IoU=0.5 2.16 #4 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD Zero-Shot CLIP + Guidance Model R@10,IoU=0.1 24.30 #4 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD Zero-Shot CLIP + Guidance Model R@10,IoU=0.3 17.73 #4 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD Zero-Shot CLIP + Guidance Model R@10,IoU=0.5 11.09 #4 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD Zero-Shot CLIP + Guidance Model R@100,IoU=0.1 47.35 #4 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD Zero-Shot CLIP + Guidance Model R@100,IoU=0.3 39.58 #4 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD Zero-Shot CLIP + Guidance Model R@100,IoU=0.5 29.68 #4 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD Zero-Shot CLIP + Guidance Model R@5,IoU=0.1 18.96 #4 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD Zero-Shot CLIP + Guidance Model R@5,IoU=0.3 13.06 #4 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD Zero-Shot CLIP + Guidance Model R@5,IoU=0.5 7.4 #4 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD Zero-Shot CLIP + Guidance Model R@50,IoU=0.1 39.79 #4 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD Zero-Shot CLIP + Guidance Model R@50,IoU=0.3 32.23 #4 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD Zero-Shot CLIP + Guidance Model R@50,IoU=0.5 23.21 #4 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD VLG-Net + Guidance Model R@1,IoU=0.1 5.60 #6 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD VLG-Net + Guidance Model R@1,IoU=0.3 4.28 #6 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD VLG-Net + Guidance Model R@1,IoU=0.5 2.48 #6 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD VLG-Net + Guidance Model R@10,IoU=0.1 23.64 #6 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD VLG-Net + Guidance Model R@10,IoU=0.3 19.86 #6 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD VLG-Net + Guidance Model R@10,IoU=0.5 13.72 #6 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD VLG-Net + Guidance Model R@100,IoU=0.1 55.59 #6 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD VLG-Net + Guidance Model R@100,IoU=0.3 49.38 #6 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD VLG-Net + Guidance Model R@100,IoU=0.5 39.12 #6 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD VLG-Net + Guidance Model R@5,IoU=0.1 16.07 #6 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD VLG-Net + Guidance Model R@5,IoU=0.5 8.78 #6 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD VLG-Net + Guidance Model R@50,IoU=0.1 45.35 #6 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD VLG-Net + Guidance Model R@50,IoU=0.3 39.77 #6 of 8 Archive leaderboard report
Natural Language Moment Retrieval MAD VLG-Net + Guidance Model R@50,IoU=0.5 30.22 #6 of 8 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

BASEPruningfail

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections