Abstract
Pedestrian detection and semantic segmentation are highpotential tasks for many real-time applications. Howevermost of the top performing approaches provide state of artresults at high computational costs. In this work we pro-pose a fast solution for achieving state of art results for botpedestrian detection and semantic segmentation. As baseline for pedestrian detection we use sliding win-dows over cost efficient multiresolution filtered LUV+HOGchannels. We use the same channels for classifying pixelsinto eight semantic classes. Using short range and longrange multiresolution channel features we achieve morerobust segmentation results compared to traditional code-book based approaches at much lower computational costs. The resulting segmentations are used as additional seman-tic channels in order to achieve a more powerful pedestriandetector. To also achieve fast pedestrian detection we employ a multiscale detection scheme based on a single flexible pedestrian model and a single image scale. The proposed solution provides competitive results on both pedestrian detection and semantic segmentation benchmarks at 8 FPS on CPU and at 15 FPS on GPU, being the fastest top performing approach.