Designing Convolutional Neural Networks Using a Genetic Approach for Ball Detection

Felbinger, Georg Christian; Göttsch, Patrick; Loth, Pascal; Peters, Lasse; Wege, Felix

doi:10.1007/978-3-030-27544-0_12

Cited by 7 publications

(5 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Leiva et al [8] proposed a framework that detects the ball and other nao players, their orientation and key features of the field without using any colour information, as all processing is done on grayscale images and cascade methodology is used which combines classical approaches and modern CNN based classifiers. Felbinger et al [17] proposed the genetic algorithm approach, which optimized the CNN hyper parameters, with minimum size dataset that resulted in cost effective inference on nao robot.…”

Section: The Development Trend In Robocup On Object Detection and Tra...mentioning

confidence: 99%

Object detection on robosoccer environment using convolution neural network

Steffi¹,

Mehta²,

Venkatesh

2022

IJEECS

View full text Add to dashboard Cite

Robots with autonomous capabilities depend on vision capabilities to detect and interact with objects and their environment. In the field of robotic research, one of the focus areas is the robosoccer platform that is being used to implement and test new ideas and findings on computer vision and decision making. In this article, an efficient real-time object detection algorithm is employed in a robosoccer simulation environment by deploying a convolution neural network and Kalman filter based tracking algorithms. This study's objective is to classify nao, ball, and the goalpost as well as to validate nao and ball tracking without human intervention from initial frame to last frame. In comparison with the existing methods, the proposed method is robust and fast in identifying three classes namely nao, ball, and goalpost with a speed of 1.67 FPS and a mAP of 95.18%. By implementing this approach, soccer playing robots can make appropriate decisions during game play.

show abstract

Section: The Development Trend In Robocup On Object Detection and Tra...mentioning

confidence: 99%

Object detection on robosoccer environment using convolution neural network

Steffi¹,

Mehta²,

Venkatesh

2022

IJEECS

View full text Add to dashboard Cite

show abstract

“…Pada penelitian sebelumnya, metode yang sudah pernah dilakukan untuk mendeteksi bola dan gawang menggunakan kamera omnidirectional adalah pendeteksian menggunakan metode radial search line [29][30], [31]. Selain itu juga metode yang bisa digunakan juga untuk mendeteksi bola dan gawang pada kamera omnidirectional adalah dengan menggunakan metode color filtering HSV dan menggunakan metode scan lines [32]. Pada deteksi bola dan gawang, hal yang menjadi permasalahan deteksi adalah ketika objek atau robot bergerak.…”

Section: Pendahuluanunclassified

Deteksi Bola dan Gawang dengan Metode YOLO Menggunakan Kamera Omnidirectional pada Robot KRSBI-B

Sanubari

Puriyanto²

2022

Buletin Ilmiah Sarjana Teknik Elektro

View full text Add to dashboard Cite

This research is a form of development of object detection capabilities on wheeled soccer robots using an omnidirectional camera with the You Only Look Once (YOLO) method where the results show that the robot can detect more than one object, namely the ball and the goal on the green field. This study uses the KRSBI-Wheeled UAD robot using an omnidirectional camera as a tool to carry out the detection process and then uses OpenCV 4.0, Deep Learning, and a laptop as a place to create a detection model, as well as balls and goals as objects to be detected. The results obtained from this study are that the two types of YOLO models tested, namely YOLOv3 and YOLOv3-Tiny can detect ball and goal objects in two different types of frame sizes, namely 320x320 and 416x416 which can be seen from the performance of the YOLOv3 model which has an mAP value of 76%. on the 320x320 frame and an mAP value of 87.5% in the 416x416 frame then the YOLOv3-Tiny model has an mAP value of 68.1% in the 320x320 frame and an mAP value of 75.5% in the 416x416 frame where the YOLOv3 model can detect both object class is much more stable compared to YOLOv3-Tiny. Penelitian ini merupakan bentuk pengembangan dari kemampuan deteksi objek pada robot sepak bola beroda dengan menggunakan kamera omnidirectional dengan metode You Only Look Once (YOLO) dimana hasil penelitian menunjukkan bahwa robot dapat mendeteksi lebih dari satu objek yaitu bola dan gawang di atas lapangan hijau. Penelitian ini menggunakan robot KRSBI-Beroda UAD dengan memakai kamera omnidirectional sebagai alat untuk melakukan proses pendeteksian lalu menggunakan OpenCV 4.0, Deep Learning, dan laptop sebagai tempat membuat model pendeteksian, serta bola dan gawang sebagai objek yang akan dideteksi. Hasil yang didapatkan dari penelitian ini yaitu kedua jenis model YOLO yang diuji yaitu YOLOv3 dan YOLOv3-Tiny dapat mendeteksi objek bola dan gawang pada dua jenis ukuran frame yang berbeda yaitu 320x320 dan 416x416 yang dapat dilihat dari performa pada model YOLOv3 memiliki nilai mAP sebesar 76% pada frame 320x320 dan serta nilai mAP sebesar 87,5% pada frame 416x416 lalu pada model YOLOv3-Tiny memiliki nilai mAP sebesar 68,1% pada frame 320x320 serta nilai mAP sebesar 75,5% pada frame 416x416 yang dimana model YOLOv3 dapat mendeteksi kedua kelas objek jauh lebih stabil dibandingkan dengan model YOLOv3-Tiny.

show abstract

“…Utilizing the temporal information in the CNN, Kukleva et al (71) presented a system that uses spatiotemporal correlation to efficiently detect and track a soccer ball based on its trajectory. Felbinger et al (72) designed a CNN for ball detection using a genetic approach that optimized network hyperparameters, providing a costeffective inference on the NAO with a limited amount of training data.…”

Section: Visionmentioning

confidence: 99%

Scientific and Technological Challenges in RoboCup

Asada

Stryk

2020

Annu. Rev. Control Robot. Auton. Syst.

View full text Add to dashboard Cite

Since its inception in 1997, RoboCup has developed into a truly unique and long-standing research community advancing robotics and artificial intelligence through various challenges, benchmarks, and test fields. The main purposes of this article are to evaluate the research and development achievements so far and to identify new challenges and related new research issues. Unlike other robot competitions and research conferences, RoboCup eliminates the boundaries between pure research activities and the development of full system designs with hardware and software implementations at a site open to the public. It also creates specific scientific and technological research and development challenges to be addressed. In this article, we provide an overview of RoboCup, including its league structure and related research issues. We also review recent studies across several research categories to show how participants (called RoboCuppers) address the research and development challenges before, during, and after the annual competitions. Among the diversity of research issues, we highlight two unique aspects of the challenges: the platform design of the robots and the game evaluations. Both of these aspects contribute to solving the research and development challenges of RoboCup and verifying the results from a common perspective (i.e., a more objective view). Finally, we provide concluding remarks and discuss future research directions.

show abstract

Designing Convolutional Neural Networks Using a Genetic Approach for Ball Detection

Cited by 7 publications

References 5 publications

Object detection on robosoccer environment using convolution neural network

Object detection on robosoccer environment using convolution neural network

Deteksi Bola dan Gawang dengan Metode YOLO Menggunakan Kamera Omnidirectional pada Robot KRSBI-B

Scientific and Technological Challenges in RoboCup

Contact Info

Product

Resources

About