Abstract. A new algorithm for the incremental learning and non-intrusive tracking of the appearance of a previously non-seen face is presented. The computation is done in a causal fashion: the information for a given frame to be processed is combined only with the one of previous frames. To achieve this aim, a novel way for simultaneous and incremental computation of the Singular Value Decomposition (SVD) and the mean of the data is explained in this work. Previous developed methods about computing the SVD iteratively are taken into account and a novel way to extract the mean from a factorised matrix using SVD is obtained. Moreover, the results are achieved with linear computational cost and sublinear memory requirements with respect to the size of the data. Some experimental results are also reported. 1

segmentation refinement and improve ro-bustness to noise. Experiments show that our hand detec-tion technique based on the proposed on-the-fly training ap-proach significantly outperforms state-of-the-art techniques with respect to accuracy and robustness on a variety of chal-lenging videos. This is due

-dimensional graph searching problem. DP provides mathematically optimal boundaries while greatly reducing sensitivity to local noise or other intervening structures. Robustness is further enhanced with on-the-fly training which causes the boundary to adhere to the specific type of edge currently being followed

enhancements to the basic live wire methodology include boundary cooling and on-the-fly training. Data-driven boundary cooling generates seed points automatically and further reduces user input. On-the-fly training adapts the dynamic boundary to edges of current interest. Using the live wire technique

Recently SVMs using spatial pyramid matching (SPM) kernel have been highly successful in image classification. Despite its popularity, these nonlinear SVMs have a complexity O(n 2 ∼ n 3) in training and O(n) in testing, where n is the training size, implying that it is nontrivial to scaleup

input depth map & 2D features data-driven tracking our tracking data-driven retargeting our retargeting Figure 1: Our adaptive tracking model conforms to the input expressions on-the-fly, producing a better fit to the user than state-of-the-art data driven techniques [Weise et al. 2011] which

in a weighted graph. Optimal graph searching provides mathematically piece-wise optimal bound-aries while greatly reducing sensitivity to local noise or other intervening structures. Robustness is further enhanced with on-the-fly training which causes the boundary to adhere to the specific type of edge

on offline training makes the approach inconvenient to adopt and hard to apply to many practical FSM applications, which often deal with a large variety of inputs different from train-ing inputs. This work presents an assembly of techniques that com-pletely remove the needs for offline training

Spoken language systems often rely on static speech recognizers. When the underlying models are improved on-the-fly, training is usually performed using unsupervised methods. In this work, we explore an alternative approach that uses human computation to provide crowd-supervised training of a

, evaluate outcomes, and modify parameters in real-time within a single software environment. We describe our meta-instrument, the Wekinator, which allows a user to engage in on-the-fly learning using arbitrary control modalities and sound synthesis environments. We provide details regarding the system