Deep Forest with Local Experts Based on ELM for Pedestrian Detection

Despite recent significant advances, pedestrian detection continues to be an extremely challenging problem in real scenarios. Recently, some authors have shown the advantages of using combinations of part/patch-based detectors in order to cope with the large variability of poses and the existence of partial occlusions. In the beginning of 2017, deep forest is put forward to make up the blank of the decision tree in the field of deep learning. Deep forests have much less parameters than deep neural network and the advantages of higher classification accuracy. In this paper, we propose a novel pedestrian detection approach that combines the flexibility of a part-based model with the fast execution time of a deep forest classifier. In this proposed combination, the role of the part evaluations is taken over by local expert evaluations at the nodes of the decision tree. We first do feature select based on extreme learning machines to get feature sets. Afterwards we use the deep forest to classify the feature sets to get the score which is the results of the local experts. We tested the proposed method with well-known challenging datasets such as TUD and INRIA. The final experimental results on two challenging pedestrian datasets indicate that the proposed method achieves the state-of-the-art or competitive performance.