Part-Aware Interactive Learning for Scene Graph Generation

Tian, Hongshuo; Xu, Ning; Liu, An-An; Zhang, Yongdong

doi:10.1145/3394171.3413501

Cited by 12 publications

(3 citation statements)

References 39 publications

(60 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…When modeling the interaction of inter-actor parts, we need to explore the mutually important and relevant features among parts of different individuals. To solve this problem, we learn from the mutual attention mechanism [ 47 , 48 ], which is originally introduced in social networks. The mutual attention mechanism is also used to calculate the edge weight vector in the visual graph in our work.…”

Section: Methodsmentioning

confidence: 99%

Multi-Perspective Representation to Part-Based Graph for Group Activity Recognition

Lang

Xiang

et al. 2022

Sensors

View full text Add to dashboard Cite

Group activity recognition that infers the activity of a group of people is a challenging task and has received a great deal of interest in recent years. Different from individual action recognition, group activity recognition needs to model not only the visual cues of individuals but also the relationships between them. The existing approaches inferred relations based on the holistic features of the individual. However, parts of the human body, such as the head, hands, legs, and their relationships, are the critical cues in most group activities. In this paper, we establish the part-based graphs from different viewpoints. The intra-actor part graph is designed to model the spatial relations of different parts for an individual, and the inter-actor part graph is proposed to explore part-level relations among actors, in which visual relation and location relation are both considered. Furthermore, a two-branch framework is utilized to capture the static spatial and dynamic temporal representations simultaneously. On the Volleyball Dataset, our approach obtains a classification accuracy of 94.8%, achieving very competitive performance in comparison with the state of the art. As for the Collective Activity Dataset, our approach improves the accuracy by 0.3% compared with the state-of-the-art results.

show abstract

Section: Methodsmentioning

confidence: 99%

Multi-Perspective Representation to Part-Based Graph for Group Activity Recognition

Lang

Xiang

et al. 2022

Sensors

View full text Add to dashboard Cite

show abstract

“…They used the Naive Bayes classifier, K-Nearest Neighbor Algorithm, Back Propagation Neural Network Algorithm, C4.5, and Support Vector Machines as classification techniques. A proposed cost-effective mobile phone-based device solution that is both cost-effective and noise-resistant (Tian et al 2021). They used features such as Local Binary Pattern (LBP), Gabor, and Histogram-based features to differentiate between different types of obstacles.…”

Section: Literature Reviewmentioning

confidence: 99%

Auxiliary Variables and Two-Step Iterative Algorithms in Object Detection: A Novel Approach

Rahman,

Reza,

Tanvir

et al. 2023

Proceedings of the International Conference on Industrial Engineering and Operations Management

View full text Add to dashboard Cite

This research aims to improve the quality of life for visually impaired individuals by identifying obstacles in an indoor environment using a solution approach. The approach uses a Convolutional Neural Network (CNN) to extract features and detect objects from real-time video. A head-mounted image acquisition device detects objects and provides audio information to the visually impaired. The system processes live video streams frame-by-frame, processing each frame as separate images. The authors incorporate additional variables into a novel two-step iterative approach to solve complex computer vision problems. The proposed method improves computational efficiency and accuracy, opening up promising avenues for future research. The paper also presents a new mathematical formulation of curve and surface reconstitute algorithms by introducing auxiliary variables. Moving object detection is crucial for intelligent video monitoring systems, as it allows for accurate extraction of foreground objects. This paper aims to detect real moving objects from un-stationary backgrounds, limiting false negatives and achieving maximum application independence. The model and motion of the target objects are assumed to be unknown. Introduce a new mathematical formulation for curve and surface reconstruction algorithms, introducing auxiliary variables and minimizing energy. This approach transforms an implicit data constraint into an explicit convex reconstruction problem, simplifying it. The formulation also allows for more precise parameter settings, ensuring convergence to a minimum. And demonstrate the properties and results of this new auxiliary problem, mainly when the potential is a function of the distance to the closest feature point.

show abstract

“…For understanding beyond standard labels, several papers have investigated the automatic learning and decomposition of tasks from instructional videos [1,39,53]. Other work seeks to uncover the scene graph and the interactions of a video, where nodes denote objects and edges denote activities [26,31,36,58,63,66,68]. We also aim to learn from long videos beyond standard labels.…”

Section: Multimodal Video Understandingmentioning

confidence: 99%