人工智能培训

搜索

人工智能教程:一种探测自然图像人体视觉分割的理想观测器模型(An Ideal Observer Model to Probe Human Visual Segm

[复制链接]
mxs810 发表于 2018-6-4 09:01:28 | 显示全部楼层 |阅读模式
mxs810 2018-6-4 09:01:28 380 0 显示全部楼层
人工智能教程:一种探测自然图像人体视觉分割的理想观测器模型(An Ideal Observer Model to Probe Human Visual Segmentation of Natural  Images)视觉分割是一种关键的感知功能,它划分视觉空间并允许在复杂环境中检测,识别和区分对象。人类对自然图像进行分割的过程目前仍然知之甚少。现有的数据集依赖于手动标记来描述感知,运动和认知因素。部分原因在于,我们缺乏理想的分割观察者模型来指导受约束的实验。另一方面,尽管机器学习最近取得了进展,现代算法仍然没有达到人类细分的性能。我们的目标有两个:(1)提出一个模型来探测人类视觉分割机制;(2)开发一个有效的图像分割算法。为此,我们提出了一种新颖的视觉分割概率生成模型,它首次将1)关于视觉皮层中神经元的敏感性的知识与自然图像中的统计规律相结合;和2)分割图(即视觉空间的分割)上的非参数贝叶斯先验。我们提供一种学习和推理算法,验证合成数据,并说明我们模型的两个组成部分如何改善自然图像的分割。然后我们表明,后部分布式分割很好地捕捉了人类主体之间的变异性,表明我们的模型提供了探索人类视觉分割的可行方法。
Visual segmentation is a key perceptual function that partitions visual spaceand allows for detection, recognition and discrimination of objects in complexenvironments.The processes underlying human segmentation of natural images arestill poorly understood.Existing datasets rely on manual labeling thatconflate perceptual, motor, and cognitive factors.In part, this is because welack an ideal observer model of segmentation to guide constrained experiments.On the other hand, despite recent progress in machine learning, modernalgorithms still fall short of human segmentation performance.Our goal here istwo-fold (i) propose a model to probe human visual segmentation mechanisms and(ii) develop an efficient algorithm for image segmentation.To this aim, wepropose a novel probabilistic generative model of visual segmentation that forthe first time combines 1) knowledge about the sensitivity of neurons in thevisual cortex to statistical regularities in natural images;and 2)non-parametric Bayesian priors over segmentation maps (ie partitions of thevisual space).We provide an algorithm for learning and inference, validate iton synthetic data, and illustrate how the two components of our model improvesegmentation of natural images.We then show that the posterior distributionover segmentations captures well the variability across human subjects,indicating that our model provides a viable approach to probe human visualsegmentation.人工智能教程:一种探测自然图像人体视觉分割的理想观测器模型(An Ideal Observer Model to Probe Human Visual Segmentation of Natural  Images) KIxXbg08R02daX9K.jpg
URL地址:https://arxiv.org/abs/1806.00111     ----pdf下载地址:http://arxiv.org/pdf/1806.00111    ----人工智能教程:一种探测自然图像人体视觉分割的理想观测器模型(An Ideal Observer Model to Probe Human Visual Segmentation of Natural  Images)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

mxs810当前离线
新手上路

查看:380 | 回复:0

快速回复 返回顶部 返回列表