PSC-Net: learning part spatial co-occurrence for occluded pedestrian detection

Jin Xie, Tianjin University
Yanwei Pang, Tianjin University
Hisham Cholakkal, Inception Institute of Artificial Intelligence
Rao Anwer, Inception Institute of Artificial Intelligence
Fahad Khan, Inception Institute of Artificial Intelligence
Ling Shao, Inception Institute of Artificial Intelligence

IR Deposit conditions:

  • OA version (pathway b)
  • Accepted version
  • 12 month embargo
  • Published source must be acknowledged
  • Must link to publisher version with DOI

Abstract

Detecting pedestrians, especially under heavy occlusion, is a challenging computer vision problem with numerous real-world applications. This paper introduces a novel approach, termed as PSC-Net, for occluded pedestrian detection. The proposed PSC-Net contains a dedicated module that is designed to explicitly capture both inter and intra-part co-occurrence information of different pedestrian body parts through a graph convolutional network (GCN). Both inter and intra-part co-occurrence information contribute towards improving the feature representation for handling varying level of occlusions, ranging from partial to severe occlusions. Our PSC-Net exploits the topological structure of pedestrian and does not require part-based annotations or additional visible bounding-box (VBB) information to learn part spatial co-occurrence. Comprehensive experiments are performed on three challenging datasets: CityPersons, Caltech, and CrowdHuman datasets. Particularly, in terms of log-average miss rates and with the same backbone and input scale as those of the state-of-the-art MGAN, the proposed PSC-Net achieves absolute gains of 4.0% and 3.4% over MGAN on the heavy occlusion subsets of CityPersons and Caltech test sets, respectively.