This paper addresses the problem of event detection and localization in long football (soccer) videos. Our key idea is that understanding the long-range dependencies between video frames is imperative for accurate event localization in long football videos. Additionally, proper event detection is not likely for fast movements in football videos without considering mid-range and short-range correlations between neighboring video frames. We argue that event spotting can be considerably improved by considering short-range to long-range frame dependencies in a unified architecture. To model long-range and midrange dependencies, we propose to use the dilated recurrent neural network (DilatedRNN) with long shortterm memory (LSTM) units, grounded on two-stream convolutional neural network (Two-stream CNN) features. While two-stream CNN extracts local spatiotemporal features necessary for fine-level details, the DilatedRNN makes the information obtained from distant frames available for the classifier and spotting algorithms. Evaluating our event spotting algorithm on the largest publicly available benchmark football dataset -SoccerNet-shows an accuracy improvement of 0.8% -13.6% compared to state of the art, and up to 30.1% accuracy gain in comparison to the baselines. We also investigate the contribution of each neural network component in spotting accuracy through an extensive ablation study.INDEX TERMS Deep learning, Dilated RNNs, Sport videos event detection, Two-stream CNNs.
This paper presents a development of a novel path planning algorithm, called Generalized Laser simulator (GLS), for solving the mobile robot path planning problem in a two-dimensional map with the presence of constraints. This approach gives the possibility to find the path for a wheel mobile robot considering some constraints during the robot movement in both known and unknown environments. The feasible path is determined between the start and goal positions by generating wave of points in all direction towards the goal point with adhering to constraints. In simulation, the proposed method has been tested in several working environments with different degrees of complexity. The results demonstrated that the proposed method is able to generate efficiently an optimal collision-free path. Moreover, the performance of the proposed method was compared with the A-star and laser simulator (LS) algorithms in terms of path length, computational time and path smoothness. The results revealed that the proposed method has shortest path length, less computational time and the best smooth path. As an average, GLS is faster than A * and LS by 7.8 and 5.5 times, respectively and presents a path shorter than A * and LS by 1.2 and 1.5 times. In order to verify the performance of the developed method in dealing with constraints, an experimental study was carried out using a Wheeled Mobile Robot (WMR) platform in labs and roads. The experimental work investigates a complete autonomous WMR path planning in the lab and road environments using a live video streaming. Local maps were built using data from a live 2698 CMC, 2022, vol.71, no.2 video streaming with real-time image processing to detect segments of the analogous-road in lab or real-road environments. The study shows that the proposed method is able to generate shortest path and best smooth trajectory from start to goal points in comparison with laser simulator.
This paper describes the development of a mid-line detection system on curve road as guidance for drivers to stay center in road-lane they are currently on using simulation model. The system will identify the curve road and detecting the tangent for each segmented curve. The purpose of detecting the tangent is to find a point that is normal to tangent of the curve. Then, using pixel distance calculation, we can get a midpoint for the curve road in order to detect and draw a virtual mid-line. This mid-line will be the guidance for the drivers to stay center when driving on the curve road. As a safety measure, the system will notify the driver with a warning message if the vehicle goes off the lane. Yet, if the driver decided to change lane, the system will automatically update the new detection on left and right boundaries. The warning message will turn off once it gets back on track. In this paper, we used B-spline as a base to study the curve behavior from simple to complicated curves. As for the method to measure the curve, we combine several algorithms from B-spline together with Generalised Hough Transform to determine the transformation parameter and the position of the model in the image. The proposed method gives a unified framework for detecting, refining and tracking the road lane. Experimental result using images of real road scene are presented.
The use of graphics has a tendency to aid reasoning in program solving by improving novice programmers’ ability to read and write code. This study extends existing work in computer programming on the use of diagrammatic representation for students undertaking the fundamental data structure course (CS2) in Malaysia. Students were tested on comprehension of diagrams followed by the composition of code with respect to the linked list topic. The data was assessed using the inter-rater agreement test and showed a high degree of consistent ratings. Results showed a moderate correlation between students’ ability to analyze list operations in the form of notation and performance on code writing. Students assessed the diagrams differently according to the complexity level. The result can be generalized to conclude that the use of diagrams alone may not fully support reasoning and program solving. However, some types of diagrams are potentially more effective to support code composition and more emphasis should be given to evaluating the effectiveness of diagrams in organizing cues to facilitate novice programmers in program solving. Further investigation on a combination of activities related to comprehension of diagrams, including code reading and explanation prior to code writing, is recommended.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.