Projects

Phase 2 Projects

Efficient active online learning for 3D reconstruction and scene understanding


Project leaders: Daniel Cremers (Garching), Rudolph Triebel (Garching)

Researchers:

Administration: Sabine Wagner (Garching)

Associates:

Summary:

Machine learning algorithms have become an important building block in modern  computer vision systems. However, most of these systems are based on offline learning, i.e. learning is done only once before system deployment, and future adaptations to unobserved circumstances are not considered. Furthermore, the amount of labelled training data required for the learning task is usually very high, because the system has no possibility to select subsets of the training data that are particularly suited for the required classification task. Quite in contrast, human perception strongly relies on our capacity to constantly learn and adapt our acquired knowledge to new environments and circumstances.
The goal of this project is to address this challenge by developing novel learning methods that perform the learning task with a higher degree of autonomy than current state- of-the-art systems. Here, autonomy refers to the ability of the system to decide which kind of information is more useful for more efficient learning. To achieve this, we will develop an Active Learning system, in which learning is done in cycles: After an initial training phase the system is presented with new observations for classification. Among them it selects those that are most informative. For these, labels are queried from a  human and the so obtained labeled data are used in the next round of training. This will  lead to two major improvements over current systems: first, the required amount of  labelled training data will be significantly lower, because the used training data will be  much more context-oriented. And second, the system will have a much higher capability  to adapt to new situations or environments, because learning is done in an ongoing  process. Our Active Learning system will be applied to two important research challenges  in computer vision, namely 3D reconstruction and scene understanding, and  the aim is to achieve substantial performance improvements in these areas using our  proposed autonomous, i.e. in this case active learning approach.

Phase 1 Projects

Efficient active online learning for 3D reconstruction and scene understanding


Project leaders: Daniel Cremers (Garching), Rudolph Triebel (Garching)

Researchers:

Administration: Sabine Wagner (Garching)

Associates:

Summary:

Machine learning algorithms have become an important building block in modern  computer vision systems. However, most of these systems are based on offline learning, i.e. learning is done only once before system deployment, and future adaptations to unobserved circumstances are not considered. Furthermore, the amount of labelled training data required for the learning task is usually very high, because the system has no possibility to select subsets of the training data that are particularly suited for the required classification task. Quite in contrast, human perception strongly relies on our capacity to constantly learn and adapt our acquired knowledge to new environments and circumstances.
The goal of this project is to address this challenge by developing novel learning methods that perform the learning task with a higher degree of autonomy than current state- of-the-art systems. Here, autonomy refers to the ability of the system to decide which kind of information is more useful for more efficient learning. To achieve this, we will develop an Active Learning system, in which learning is done in cycles: After an initial training phase the system is presented with new observations for classification. Among them it selects those that are most informative. For these, labels are queried from a  human and the so obtained labeled data are used in the next round of training. This will  lead to two major improvements over current systems: first, the required amount of  labelled training data will be significantly lower, because the used training data will be  much more context-oriented. And second, the system will have a much higher capability  to adapt to new situations or environments, because learning is done in an ongoing  process. Our Active Learning system will be applied to two important research challenges  in computer vision, namely 3D reconstruction and scene understanding, and  the aim is to achieve substantial performance improvements in these areas using our  proposed autonomous, i.e. in this case active learning approach.