Pedestrian detection + a convolutional network model (with a few twists).
1、Unsupervised convolutional sparse auto-encoders to pre-train features at all levels from the relatively small INNRIA dataset.
2、End-to-end supervised training to train the classifier and fine-tune the features in an intregrated fashion.
3、Multi-stage features with layer-skipping connections enable output stages to combine global shape detectors with local motif detectors.