Boosted Deep Neural Network with Weighted Output Layers 2017-01-1997
Vision based driving environment perception is current research hotspot in automatic driving field, which has made great progress due to the continuous breakthroughs in the research of deep neural network. As is well known, deep neural network has won tremendous successes in a wide variety of image recognition tasks, such as pedestrian detection and vehicle identification, which have accomplished the commercialization successfully in intelligent monitor system. Nevertheless, driving environment perception has a higher request for the generalization performance of deep neural network, which needs further studies on its design and training methods.
In this paper, we presented a new boosted deep neural network in order to improve its generalization performance and meanwhile keep computational budget constant. Above all, the most representative methods to improve the generalization performance of deep neural network were introduced. Next, we analyzed the merits and demerits of these methods under limited training samples and computation resources. Then we described a new boosted deep neural network with weighted output layers. On one hand, there are several output layers that constitute sequential classifiers, which boost the final performance of presented deep neural network. On the other hand, it saves the computation consumption through sharing partial network structure among the classifiers. Our proposed model improves the generalization performance and avoids excessively increasing computing at the same time. Finally, we made experiments to confirm the effectiveness of our model.