Abstract
In this paper, a new region-level scene labeling approach is proposed, which combines a deep convolutional neural network with the Mean Shift segmentation algorithm. For each image, it is firstly segmented into local regions using the Mean Shift algorithm. Then a deep convolutional neural network trained with images of target objects is employed to get the probability scores of randomly cropped samples of each segment. Object category of each local segment is finally determined by the average probability scores of its samples. This method alleviates the need for hand-crafted features, and produces a powerful representation that captures texture, shape, and contextual information. Experiment results in a campus environment have demonstrated that the proposed method can achieve satisfactory labeling accuracy.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have