Visual Mesh: Real-Time Object Detection Using Constant Sample Density

Houliston, Trent; Chalup, Stephan K.

doi:10.1007/978-3-030-27544-0_4

Cited by 10 publications

(7 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Ball-only CNNs had input size massively reduced to be ported to typical robots for the humanoid league [32]. Even the so-called Visual Mesh technique [33] that uses multiple scales to improve the performance of neural networks requires a method to propose sub-regions. The classifier performs a crucial role in proposing a sub-region.…”

Section: Human Vision and Computer Visionmentioning

confidence: 99%

Human-In-The-Loop Construction of Decision Tree Classifiers with Parallel Coordinates

Estivill‐Castro

Gilmore

Hexel

2020

2020 IEEE International Conference on Systems, Man, and Cybernetics (SMC)

View full text Add to dashboard Cite

How can there be Human-In-the-Loop-Learning (HILL) if datasets aimed at building classifiers have ever more dimensions? We make two contributions. First, we examine the few early results on the effectiveness of HILL for building autonomous classifiers and report on our own experiment that validates the merits of HILL. Second, we introduce a HILL system (by using parallel coordinates) for learning of decision tree classifiers (DTCs). DTCs importantly emphasise the relevance of attributes and enable attribute selection, and therefore are appreciated for their transparency.The proposed system addresses a number of the shortcomings of the many HILL systems and allows for easy exploration of datasets. In particular, we incorporate parallel coordinates effectively in our tool for visualisation of high dimensional datasets. We can not only focus the learning on the accuracy of classifiers, but we can enhance performance in other important factors such as system's interpretability and the ability to gain insight into datasets. Finally, we show the advantages of our HILL system in the application area of mobile robotics using the case study of image segmentation in robotic soccer.

show abstract

Section: Human Vision and Computer Visionmentioning

confidence: 99%

Human-In-The-Loop Construction of Decision Tree Classifiers with Parallel Coordinates

Estivill‐Castro

Gilmore

Hexel

2020

2020 IEEE International Conference on Systems, Man, and Cybernetics (SMC)

View full text Add to dashboard Cite

show abstract

“…Although, this network achieved some good results in near real-time performance on a low-power processor, it trained and evaluated on a limited data set that contains only one ball type. A real-time CNN-based object detection approach for resource-constrained robotics is presented in [8]. Before feeding the input image to the network, it transforms the image using the object geometry to form a Visual Mesh.…”

Section: Related Workmentioning

confidence: 99%

A Real-Time Ball Detection Approach Using Convolutional Neural Networks

Teimouri

Delavaran

Rezaei

2019

RoboCup 2019: Robot World Cup XXIII

View full text Add to dashboard Cite

Ball detection is one of the most important tasks in the context of soccer-playing robots. The ball is a small moving object which can be blurred and occluded in many situations. Several neural network based methods with different architectures are proposed to deal with the ball detection. However, they are either neglecting to consider the computationally low resources of humanoid robots or highly depend on manually-tuned heuristic methods to extract the ball candidates. In this paper, we propose a new ball detection method for low-cost humanoid robots that can detect most soccer balls with a high accuracy rate of up to 97.17%. The proposed method is divided into two steps. First, some coarse regions that may contain a full ball are extracted using an iterative method employing an efficient integral image based feature. Then they are fed to a lightweight convolutional neural network to finalize the bounding box of a ball. We have evaluated the proposed approach using a comprehensive dataset and the experimental results show the efficiency of our method.

show abstract

“…Several teams have used convolutional neural networks either for binary classification tasks [15,2] or to detect several relevant object categories [17,13]. Houliston and Chalup [8] used a technique called Visual Mesh to improve the performance of neural networks at multiple scales. These methods, however, use CNNs for classification only, therefore they still require a separate object proposal method, and the quality of the system may largely depend on the efficiency of the algorithm used to generate candidates for classification.…”

Section: Computer Vision In Robot Soccermentioning

confidence: 99%

ROBO: Robust, Fully Neural Object Detection for Robot Soccer

Szemenyei

Estivill‐Castro

2019

RoboCup 2019: Robot World Cup XXIII

View full text Add to dashboard Cite

Deep Learning has become exceptionally popular in the last few years due to its success in computer vision [11,21,18] and other fields of AI [25,28,5]. However, deep neural networks are computationally expensive, which limits their application in low power embedded systems, such as mobile robots. In this paper, an efficient neural network architecture is proposed for the problem of detecting relevant objects in robot soccer environments. The ROBO model's increase in efficiency is achieved by exploiting the peculiarities of the environment. Compared to the state-of-the-art Tiny YOLO model, the proposed network provides approximately 35 times decrease in run time, while achieving superior average precision, although at the cost of slightly worse localization accuracy.

show abstract

Visual Mesh: Real-Time Object Detection Using Constant Sample Density

Cited by 10 publications

References 7 publications

Human-In-The-Loop Construction of Decision Tree Classifiers with Parallel Coordinates

Human-In-The-Loop Construction of Decision Tree Classifiers with Parallel Coordinates

A Real-Time Ball Detection Approach Using Convolutional Neural Networks

ROBO: Robust, Fully Neural Object Detection for Robot Soccer

Contact Info

Product

Resources

About