Deep Reinforcement Learning for a Humanoid Robot Soccer Player

Silva, Isaac Jesus da; Perico, Danilo H.; Homem, Thiago Pedro Donadon; Bianchi, Reinaldo A. C.

doi:10.1007/s10846-021-01333-1

Cited by 10 publications

(1 citation statement)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In recent years, deep neural networks have achieved great success in the field of image classification, and gesture recognition based on deep learning has become the mainstream direction [11]. Convolutional neural networks in deep learning can learn the features of input information layer by layer and classify it more accurately [12]. Oyedotum et al [13] first performed a binarization preprocessing operation on the image, then set a threshold to localize the gesture, and finally used a convolutional neural network to recognize the gesture.…”

Section: Introductionmentioning

confidence: 99%

Medical Gesture Recognition Method Based on Improved Lightweight Network

Wang

et al. 2022

Applied Sciences

View full text Add to dashboard Cite

Surgery is a compelling application field for collaborative control robots. This paper proposes a gesture recognition method applied to a medical assistant robot delivering instruments to collaborate with surgeons to complete surgeries. The key to assisting the surgeon in passing instruments in the operating room is the ability to recognize the surgeon’s hand gestures accurately and quickly. Existing gesture recognition techniques suffer from poor recognition accuracy and low rate. To address the existing shortcomings, we propose an improved lightweight convolutional neural network called E-MobileNetv2. The ECA module is added to the original MobileNetv2 network model to obtain more useful features by computing the information interactions between the current channel and the adjacent channels and between the current channel and the distant channels in the feature map. We add R6-SELU activation function to enhance the network’s ability to extract features. By adjusting the shrinkable hyper-parameters, the number of parameters of the network is reduced to improve the recognition speed. The improved network model achieves excellent performance on both the self-built dataset Gesture_II and the public dataset Jester. The recognition accuracy of the improved model is 96.82%, which is 3.17 % higher than that of the original model, achieving an increase in accuracy and recognition speed.

show abstract

Section: Introductionmentioning

confidence: 99%