Contour detection is an important and fundamental problem in computer vision that finds numerous applications. In this paper, we propose a learning algorithm for contour detection via random forest. Visual cues that can be extracted easily and efficiently are integrated to learn a detector where the decision of an contour pixel is made independently via the random forest at each location in the image. We evaluate the proposed algorithm against leading methods in the literature on the Berkeley Segmentation Dataset. Experimental results demonstrate that the proposed contour detection algorithm performs favorably against state-of-the-art methods in terms of speed and accuracy.