GRAZE: Grounded Refinement and Motion-Aware Zero-Shot Event Localization

• URL: http://arxiv.org/abs/2604.01383v1
• Date: Wed, 01 Apr 2026 20:41:58 GMT
• Title: GRAZE: Grounded Refinement and Motion-Aware Zero-Shot Event Localization
• Abstract summary: We present a training-free pipeline for FPOC localization that requires no labeled tackle-trimmed examples.<n>On 738 tackle-practice videos, GRAZE produces valid outputs for 97.4% of clips and localizes FPOC within $pm$ 10 frames on 77.5% of all clips and within $pm$ 20 frames on 82.7% of all clips.
• Score: 0.4306143768014156
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: American football practice generates video at scale, yet the interaction of interest occupies only a brief window of each long, untrimmed clip. Reliable biomechanical analysis, therefore, depends on spatiotemporal localization that identifies both the interacting entities and the onset of contact. We study First Point of Contact (FPOC), defined as the first frame in which a player physically touches a tackle dummy, in unconstrained practice footage with camera motion, clutter, multiple similarly equipped athletes, and rapid pose changes around impact. We present GRAZE, a training-free pipeline for FPOC localization that requires no labeled tackle-contact examples. GRAZE uses Grounding DINO to discover candidate player-dummy interactions, refines them with motion-aware temporal reasoning, and uses SAM2 as an explicit pixel-level verifier of contact rather than relying on detection confidence alone. This separation between candidate discovery and contact confirmation makes the approach robust to cluttered scenes and unstable grounding near impact. On 738 tackle-practice videos, GRAZE produces valid outputs for 97.4% of clips and localizes FPOC within $\pm$ 10 frames on 77.5% of all clips and within $\pm$ 20 frames on 82.7% of all clips. These results show that frame-accurate contact onset localization in real-world practice footage is feasible without task-specific training.

Related papers

• CAS-FD: Contact-Aware Temporal Sampling for Single-View Foul vs Dive Recognition [2.6686511750004915]
  We introduce a balanced 600-clip single-view Foul/Dive dataset.<n>We show that contact-aware sampling concentrating the model's attention around the moment of physical contact yields substantially improved recognition.
  arXiv  Detail & Related papers  (2026-08-17T19:06:42Z)
• SCOPE: Simulating Cross-game Operations in Playable Environments for FPS World Models [49.15128236103093]
  Existing methods inject actions globally and train on single titles, failing under dense FPS inputs.<n>We propose SCOPE, which inserts a conditioning module into each transformer block of a pretrained video diffusion model.<n>We also introduce CrossFPS, the first multi-game FPS dataset with frame-aligned action telemetry.
  arXiv  Detail & Related papers  (2026-05-22T08:06:37Z)
• SoccerLens: Grounded Soccer Video Understanding Beyond Accuracy [52.09081885158698]
  Vision-language models (VLMs) have recently shown strong potential in soccer video understanding.<n>Existing evaluation protocols focus on classification accuracy and do not assess visual grounding.<n>We introduce SoccerLens, a benchmark for grounded soccer video understanding.
  arXiv  Detail & Related papers  (2026-05-10T15:18:08Z)
• Predicting Penalty Kick Direction Using Multi-Modal Deep Learning with Pose-Guided Attention [0.0]
  This study introduces a real-time, multi-modal deep learning framework to predict the direction of a penalty kick before ball contact.<n>A custom dataset of 755 penalty kick events was created from real match videos, with frame-level annotations for object detection, shooter keypoints, and final ball placement.<n>The model achieved 89% accuracy on a held-out test set, outperforming visual-only and pose-only baselines by 14-22%.
  arXiv  Detail & Related papers  (2025-09-30T11:02:59Z)
• KASportsFormer: Kinematic Anatomy Enhanced Transformer for 3D Human Pose Estimation on Short Sports Scene Video [4.653030985708889]
  We introduce KASportsFormer, a novel transformer based 3D pose estimation framework for sports.<n>Our proposed method achieves state-of-the-art results with MPJPE errors of 58.0mm and 34.3mm, respectively.
  arXiv  Detail & Related papers  (2025-07-28T12:17:40Z)
• FACTS: Fine-Grained Action Classification for Tactical Sports [4.810476621219244]
  Classifying fine-grained actions in fast-paced, close-combat sports such as fencing and boxing presents unique challenges.<n>We introduce FACTS, a novel approach for fine-grained action recognition that processes raw video data directly.<n>Our findings enhance training, performance analysis, and spectator engagement, setting a new benchmark for action classification in tactical sports.
  arXiv  Detail & Related papers  (2024-12-21T03:00:25Z)
• PMI Sampler: Patch Similarity Guided Frame Selection for Aerial Action Recognition [52.78234467516168]
  We introduce the concept of patch mutual information (PMI) score to quantify the motion bias between adjacent frames. We present an adaptive frame selection strategy using shifted leaky ReLu and cumulative distribution function. Our method achieves a relative improvement of 2.2 - 13.8% in top-1 accuracy on UAV-Human, 6.8% on NEC Drone, and 9.0% on Diving48 datasets.
  arXiv  Detail & Related papers  (2023-04-14T00:01:11Z)
• PACE: Data-Driven Virtual Agent Interaction in Dense and Cluttered Environments [69.03289331433874]
  We present PACE, a novel method for modifying motion-captured virtual agents to interact with and move throughout dense, cluttered 3D scenes. Our approach changes a given motion sequence of a virtual agent as needed to adjust to the obstacles and objects in the environment. We compare our method with prior motion generating techniques and highlight the benefits of our method with a perceptual study and physical plausibility metrics.
  arXiv  Detail & Related papers  (2023-03-24T19:49:08Z)
• P2ANet: A Dataset and Benchmark for Dense Action Detection from Table Tennis Match Broadcasting Videos [64.57435509822416]
  This work consists of 2,721 video clips collected from the broadcasting videos of professional table tennis matches in World Table Tennis Championships and Olympiads. We formulate two sets of action detection problems -- emphaction localization and emphaction recognition. The results confirm that TheName is still a challenging task and can be used as a special benchmark for dense action detection from videos.
  arXiv  Detail & Related papers  (2022-07-26T08:34:17Z)
• Shot Contrastive Self-Supervised Learning for Scene Boundary Detection [7.4262338683231155]
  Shot contrastive learning approach (ShotCoL) learns a shot representation that maximizes the similarity between nearby shots. We show how to apply ShotCoL for the task of scene boundary detection to offer state-of-the-art performance on the MovieNet dataset. We take on the problem of finding timestamps in movies and TV episodes where video-ads can be inserted while offering a minimally disruptive viewing experience.
  arXiv  Detail & Related papers  (2021-04-28T02:35:09Z)
• Enhancing Unsupervised Video Representation Learning by Decoupling the Scene and the Motion [86.56202610716504]
  Action categories are highly related with the scene where the action happens, making the model tend to degrade to a solution where only the scene information is encoded. We propose to decouple the scene and the motion (DSM) with two simple operations, so that the model attention towards the motion information is better paid.
  arXiv  Detail & Related papers  (2020-09-12T09:54:11Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.