An official website of the United States government
We propose TubeR: a simple solution for spatio-temporal video action detection. Different from existing methods that depend on either an off-line actor detector or hand-designed actor-positional hypotheses like proposals or anchors, we propose to directly detect an action tubelet in a video by simultaneously performing action localization and recognition from a single representation. TubeR learns a set of tubelet queries and utilizes a tubelet-attention module to model the dynamic spatio-temporal nature of a video clip, which effectively reinforces the model capacity compared to using actor-positional hypotheses in the spatio-temporal space. For videos containing transitional states or scene changes, we propose a context aware classification head to utilize short-term and long-term context to strengthen action classification, and an action switch regression head for detecting the precise temporal action extent. TubeR directly produces action tubelets with variable lengths and even maintains good results for long video clips. TubeR outperforms the previous state-of-the-art on commonly used action detection datasets AVA, UCF101-24 and JHMDB51-21. Code will be available on GluonCV(https://cv.gluon.ai/).
more »
« less
Self-supervised Multi-actor Social Activity Understanding in Streaming Videos
This work addresses the problem of Social Activity Recognition (SAR), a critical component in real-world tasks like surveillance and assistive robotics. Unlike traditional event understanding approaches, SAR necessitates modeling individual actors' appearance and motions and contextualizing them within their social interactions. Traditional action localization methods fall short due to their single-actor, single-action assumption. Previous SAR research has relied heavily on densely annotated data, but privacy concerns limit their applicability in real-world settings. In this work, we propose a self-supervised approach based on multi-actor predictive learning for SAR in streaming videos. Using a visual-semantic graph structure, we model social interactions, enabling relational reasoning for robust performance with minimal labeled data. The proposed framework achieves competitive performance on standard group activity recognition benchmarks. Evaluation on three publicly available action localization benchmarks demonstrates its generalizability to arbitrary action localization.
more »
« less
- PAR ID:
- 10545618
- Publisher / Repository:
- Springer
- Date Published:
- Format(s):
- Medium: X
- Location:
- ICPR 2024
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Visual event perception tasks such as action localization have primarily focused on supervised learning settings under a static observer, i.e., the camera is static and cannot be controlled by an algorithm. They are often restricted by the quality, quantity, and diversity of annotated training data and do not often generalize to out-of-domain samples. In this work, we tackle the problem of active action localization where the goal is to localize an action while controlling the geometric and physical parameters of an active camera to keep the action in the field of view without training data. We formulate an energy-based mechanism that combines predictive learning and reactive control to perform active action localization without rewards, which can be sparse or non-existent in real-world environments. We perform extensive experiments in both simulated and real-world environments on two tasks - active object tracking and active action localization. We demonstrate that the proposed approach can generalize to different tasks and environments in a streaming fashion, without explicit rewards or training. We show that the proposed approach outperforms unsupervised baselines and obtains competitive performance compared to those trained with reinforcement learning.more » « less
-
We introduce a novel actor-critic framework that utilizes vision-language models (VLMs) and large language models (LLMs) for design concept generation, particularly for producing a diverse array of innovative solutions to a given design problem. By leveraging the extensive data repositories and pattern recognition capabilities of these models, our framework achieves this goal through enabling iterative interactions between two VLM agents: an actor (i.e., concept generator) and a critic. The actor, a custom VLM (e.g., GPT-4) created using few-shot learning and fine-tuning techniques, generates initial design concepts that are improved iteratively based on guided feedback from the critic—a prompt-engineered LLM or a set of design-specific quantitative metrics. This process aims to optimize the generated concepts with respect to four metrics: novelty, feasibility, problem–solution relevancy, and variety. The framework incorporates both long-term and short-term memory models to examine how incorporating the history of interactions impacts decision-making and concept generation outcomes. We explored the efficacy of incorporating images alongside text in conveying design ideas within our actor–critic framework by experimenting with two mediums for the agents: vision language and language only. We extensively evaluated the framework through a case study using the AskNature dataset, comparing its performance against benchmarks such as GPT-4 and real-world biomimetic designs across various industrial examples. Our findings underscore the framework’s capability to iteratively refine and enhance the initial design concepts, achieving significant improvements across all metrics. We conclude by discussing the implications of the proposed framework for various design domains, along with its limitations and several directions for future research in this domain.more » « less
-
This paper introduces a novel multimodal action recognition framework designed to address the complexities of human activity recognition in manufacturing assembly tasks through spatio-temporal knowledge fusion. Traditional unimodal and naive multimodal fusion methods often fail to capture the intricate dependencies across space, time, and modality, especially in real-world industrial environments where actions are subtle, repetitive, and context-dependent. To overcome these challenges, we propose a unified architecture that incorporates: (i) a Multi-Stage Hierarchical Reconnection module for robust spatial and temporal feature disentanglement and reintegration; (ii) a spatio-temporal regularization technique, Optimized-MixUp (OMU), that jointly augments data along spatial and temporal axes to improve generalization; and (iii) a Cross-Modal Auxiliary Feature Learning component to enhance late fusion by exploiting modality-specific complementary information. Extensive experiments conducted on four benchmark datasets, NTU RGB+D, NTU RGB+D120, HA4M, and Northwestern-UCLA, demonstrate that our method outperforms recent state-of-the-art approaches, achieving top-1 accuracy of 98.4%, 93.5%, 92.0%, and 97.3%, respectively. These results confirm the framework’s robustness, scalability, and suitability for high-precision human activity understanding in manufacturing environments. The proposed method advances the field of information fusion by offering a principled approach to integrating heterogeneous spatio-temporal data for real-world action recognition tasks.more » « less
-
Understanding the behavioral and neural dynamics of social interactions is a goal of contemporary neuroscience. Many machine learning methods have emerged in recent years to make sense of complex video and neurophysiological data that result from these experiments. Less focus has been placed on understanding how animals process acoustic information, including social vocalizations. A critical step to bridge this gap is determining the senders and receivers of acoustic information in social interactions. While sound source localization (SSL) is a classic problem in signal processing, existing approaches are limited in their ability to localize animal-generated sounds in standard laboratory environments. Advances in deep learning methods for SSL are likely to help address these limitations, however there are currently no publicly available models, datasets, or benchmarks to systematically evaluate SSL algorithms in the domain of bioacoustics. Here, we present the VCL Benchmark: the first large-scale dataset for benchmarking SSL algorithms in rodents. We acquired synchronized video and multi-channel audio recordings of 767,295 sounds with annotated ground truth sources across 9 conditions. The dataset provides benchmarks which evaluate SSL performance on real data, simulated acoustic data, and a mixture of real and simulated data. We intend for this benchmark to facilitate knowledge transfer between the neuroscience and acoustic machine learning communities, which have had limited overlap.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript
- Conference Paper:
- The DOI is not currently available.
