Convolutional neural network (CNN)

Convolutional neural network (CNN) has been proven to have the most advanced pedestrian detection accuracy. However, in the case of processing a single image that often requires billions of floating-point operations, the cost is expensive. In order to solve this complexity problem, Fast R -CNN and (you only look once, YOLO) can identify pedestrians and location information in the image, reducing the computational cost. Based on the CNN framework, a multi-spectral pedestrian detection method based on deep convolutional neural networks, using the method of fusion images, in order to better perform robust pedestrian detection, an unsupervised multi-spectral pedestrian detection method based on deep convolutional neural networks is presented. The pedestrian detection method based on spectral feature learning uses the automatic annotation framework to complement the multi-spectral data information, and iteratively annotates pedestrians in the visible channel. This method is compatible with