Learn More
In this paper, we describe a novel classification technique that separates video scenes, like office work tasks, into several scenes according to each task. Even if the difference of as a whole image frame by frame in each task is small, the difference of worker's movement is quite big due to the position of face and hands according to each task. In(More)
Recently automatic image annotation (AIA) has been arising as a key technology to support image retrieval. The representative algorithm is Semantic Multiclass Labeling (SML [1]), which constructs a parametric ge-nerative model of a distribution of local image features in a class with a gaussian mixture model. Although SML shows good accuracy, SML has not(More)
Maximum likelihood (ML) method for estimating parameters of Bayesian networks (BNs) is efficient and accurate for large samples. However, ML suffers from overfitting when the sample size is small. Bayesian methods, which are effective to avoid overfitting, have difficulties for determining optimal hyperparameters of prior distributions with good balance(More)
We propose a face recognition algorithm that utilizes novel surface features in (x, y, I(x,y)) space. A face image is considered as a surface in XYI space, and the surface is segmented into definite number of regions by using Gaussian Mixture Model. Parameters of each Gaussian distribution are determined by maximizing log-likelihood function, and stored as(More)
In this paper, we propose an adaptive learning technique for "sensor-planning", that is how to control PTZ (pan-tilt-zoom) camera sensor for human finding without prior environment information. Our idea is based on Q-learning with positive and negative rewards as optical-flow result and camera motion respectively. Additionally, an adaptive learning rate(More)