Volume : VII, Issue : II, February - 2018

A FAST OBJECT DETECTION METHOD BASED ON DEEP RESIDUAL NETWORK

Xuzheng Hao

Abstract :

 Real–time Convolution Neural Network object detection methods with limited computing devices still have in less accuracy and low detection speed. Considering the above problem, a new method based on Deep Residual Network and tiny–YOLO is proposed. The new model integrates deep residual network with pre–activation, which has more layers of convolution neural networks but less weight parameters. The experiment has been conducted on the authoritative PASCAL VOC dataset, the experimental results show that the average intersection over union is improved by 0.29% and the recall rate by 2.44% with quick detection speed, which verifies the effectiveness of the method.

Keywords :

Article: Download PDF   DOI : 10.36106/ijsr  

Cite This Article:

Xuzheng Hao, A FAST OBJECT DETECTION METHOD BASED ON DEEP RESIDUAL NETWORK, INTERNATIONAL JOURNAL OF SCIENTIFIC RESEARCH : Volume-7 | Issue-2 | February-2018

Number of Downloads : 370

References :