Supervised learning has become more and more important in a wide range of applications in Computer Vision e.g. Advanced Driver Assistance Systems (ADAS). Recently, Convolutional Neural Networks, Boosting algorithms or Random Forests have achieved state-of-the-art performance on various classification and detection tasks. These successful results have been achieved essentially using huge training datasets with ground truth labels where the bounding boxes are placed correctly. We propose a novel method to track, learn and detect dynamic objects based on very few training samples.
Required audience experience: Basic knowledge of machine learning algorithms and computer vision might be helpful.
Objective of the talk: Showing new ways to boost object detection results without huge training datasets.
Keywords: Computer Vision, CNN, Convolutional Neural Networks, Boosting, Random Forest, Driver Assistance, Automotive, Machine Learning