Deep learning boosts the quality of almost all computer vision solutions. In particular, deep learning achieves excellent performance in object detection tasks, such as pedestrian detection and traffic sign recognition.
In my talk, we will discuss how a fast and robust object detector can be built using only deep learning, without any external proposal generator. Previously, it was reported that spurious false positives can occur when neural net are densely applied in the sliding window manner. We will discuss how a cascade of neural nets and hard negatives sampling for learning can be used to reduce the number of false positives.