EVA02-AT: Egocentric Video-Language Understanding with Spatial-Temporal Rotary Positional Embeddings and Symmetric Optimization

• URL: http://arxiv.org/abs/2506.14356v1
• Date: Tue, 17 Jun 2025 09:51:51 GMT
• Title: EVA02-AT: Egocentric Video-Language Understanding with Spatial-Temporal Rotary Positional Embeddings and Symmetric Optimization
• Authors: Xiaoqi Wang, Yi Wang, Lap-Pui Chau,
• Abstract summary: We introduce EVA02-AT, a suite of EVA02-based video-language foundation models tailored to egocentric video understanding tasks.<n> EVA02-AT efficiently transfers an image-based CLIP model into a unified video encoder via a single-stage pretraining.<n>We introduce the Symmetric Multi-Similarity (SMS) loss and a novel training framework that advances all soft labels for both positive and negative pairs.
• Score: 17.622013322533423
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Egocentric video-language understanding demands both high efficiency and accurate spatial-temporal modeling. Existing approaches face three key challenges: 1) Excessive pre-training cost arising from multi-stage pre-training pipelines, 2) Ineffective spatial-temporal encoding due to manually split 3D rotary positional embeddings that hinder feature interactions, and 3) Imprecise learning objectives in soft-label multi-instance retrieval, which neglect negative pair correlations. In this paper, we introduce EVA02-AT, a suite of EVA02-based video-language foundation models tailored to egocentric video understanding tasks. EVA02-AT first efficiently transfers an image-based CLIP model into a unified video encoder via a single-stage pretraining. Second, instead of applying rotary positional embeddings to isolated dimensions, we introduce spatial-temporal rotary positional embeddings along with joint attention, which can effectively encode both spatial and temporal information on the entire hidden dimension. This joint encoding of spatial-temporal features enables the model to learn cross-axis relationships, which are crucial for accurately modeling motion and interaction in videos. Third, focusing on multi-instance video-language retrieval tasks, we introduce the Symmetric Multi-Similarity (SMS) loss and a novel training framework that advances all soft labels for both positive and negative pairs, providing a more precise learning objective. Extensive experiments on Ego4D, EPIC-Kitchens-100, and Charades-Ego under zero-shot and fine-tuning settings demonstrate that EVA02-AT achieves state-of-the-art performance across diverse egocentric video-language tasks with fewer parameters. Models with our SMS loss also show significant performance gains on multi-instance retrieval benchmarks. Our code and models are publicly available at https://github.com/xqwang14/EVA02-AT .

Related papers

• VideoMolmo: Spatio-Temporal Grounding Meets Pointing [73.25506085339252]
  VideoMolmo is a model tailored for fine-grained pointing of video sequences.<n>A novel temporal mask fusion employs SAM2 for bidirectional point propagation.<n>To evaluate the generalization of VideoMolmo, we introduce VPoMolS-temporal, a challenging out-of-distribution benchmark spanning five real-world scenarios.
  arXiv  Detail & Related papers  (2025-06-05T17:59:29Z)
• FEAT: Full-Dimensional Efficient Attention Transformer for Medical Video Generation [14.903360987684483]
  We propose FEAT, a full-dimensional efficient attention Transformer for high-quality dynamic medical videos.<n>We evaluate FEAT on standard benchmarks and downstream tasks, demonstrating that FEAT-S, with only 23% of the parameters of the state-of-the-art model Endora, achieves comparable or even superior performance.
  arXiv  Detail & Related papers  (2025-06-05T12:31:02Z)
• EgoSplat: Open-Vocabulary Egocentric Scene Understanding with Language Embedded 3D Gaussian Splatting [108.15136508964011]
  EgoSplat is a language-embedded 3D Gaussian Splatting framework for open-vocabulary egocentric scene understanding.<n>EgoSplat achieves state-of-the-art performance in both localization and segmentation tasks on two datasets.
  arXiv  Detail & Related papers  (2025-03-14T12:21:26Z)
• Coarse Correspondences Boost Spatial-Temporal Reasoning in Multimodal Language Model [51.83436609094658]
  We introduce Coarse Correspondences, a simple lightweight method that enhances MLLMs' spatial-temporal reasoning with 2D images as input. Our method uses a lightweight tracking model to identify primary object correspondences between frames in a video or across different image viewpoints. We demonstrate that this simple training-free approach brings substantial gains to GPT4-V/O consistently on four benchmarks.
  arXiv  Detail & Related papers  (2024-08-01T17:57:12Z)
• SIGMA:Sinkhorn-Guided Masked Video Modeling [69.31715194419091]
  Sinkhorn-guided Masked Video Modelling ( SIGMA) is a novel video pretraining method. We distribute features of space-time tubes evenly across a limited number of learnable clusters. Experimental results on ten datasets validate the effectiveness of SIGMA in learning more performant, temporally-aware, and robust video representations.
  arXiv  Detail & Related papers  (2024-07-22T08:04:09Z)
• Learning Fine-grained View-Invariant Representations from Unpaired Ego-Exo Videos via Temporal Alignment [71.16699226211504]
  We propose to learn fine-grained action features that are invariant to the viewpoints by aligning egocentric and exocentric videos in time. To this end, we propose AE2, a self-supervised embedding approach with two key designs. For evaluation, we establish a benchmark for fine-grained video understanding in the ego-exo context.
  arXiv  Detail & Related papers  (2023-06-08T19:54:08Z)
• Spatio-Temporal Self-Attention Network for Video Saliency Prediction [13.873682190242365]
  3D convolutional neural networks have achieved promising results for video tasks in computer vision. We propose a novel Spatio-Temporal Self-Temporal Self-Attention 3 Network (STSANet) for video saliency prediction.
  arXiv  Detail & Related papers  (2021-08-24T12:52:47Z)
• Spatial-Temporal Correlation and Topology Learning for Person Re-Identification in Videos [78.45050529204701]
  We propose a novel framework to pursue discriminative and robust representation by modeling cross-scale spatial-temporal correlation. CTL utilizes a CNN backbone and a key-points estimator to extract semantic local features from human body. It explores a context-reinforced topology to construct multi-scale graphs by considering both global contextual information and physical connections of human body.
  arXiv  Detail & Related papers  (2021-04-15T14:32:12Z)
• Comparison of Spatiotemporal Networks for Learning Video Related Tasks [0.0]
  Many methods for learning from sequences involve temporally processing 2D CNN features from the individual frames or directly utilizing 3D convolutions within high-performing 2D CNN architectures. This work constructs an MNIST-based video dataset with parameters controlling relevant facets of common video-related tasks: classification, ordering, and speed estimation. Models trained on this dataset are shown to differ in key ways depending on the task and their use of 2D convolutions, 3D convolutions, or convolutional LSTMs.
  arXiv  Detail & Related papers  (2020-09-15T19:57:50Z)

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.