Depth “prediction” using modified deep convolutional neural field

机译：深度“预测”改进的深卷积神经场

获取原文

页面导航

摘要
著录项
相似文献
相关主题

摘要

Monocular depth prediction is one of the recent research topic which offers reduced computational complexity compared to stereo depth prediction but at the cost of increased ambiguity. Most algorithms for monocular depth prediction are dependent on training models. These algorithms predict depth using learning models on semantically align databases. They leverage over scene dependent features like semantic labels, SIFT Flow. Deep CNN-CRF framework uses scene independent feature, i.e., superpixels but presence of similar superpixel in scene increases complexity. In this paper, we have proposed a segmentation based feature which is trained using modified deep CNN-CRF which results in significant reduction in complexity. We evaluated our algorithm on NYUv2 database based on error metrics. Error performance of proposed algorithm is slightly better than deep CNN-CRF but significantly better than other existing algorithms.

机译：单眼深度预测是最近的研究课题之一，与立体声深度预测相比，提供了减少的计算复杂性，但是以增加歧义的成本。用于单眼深度预测的大多数算法取决于训练模型。这些算法使用在语义对齐数据库上的学习模型预测深度。他们利用场景依赖性特征，如语义标签，筛选流量。深度CNN-CRF框架使用场景独立特征，即Superpixels，但场景中类似的Superpixel的存在会增加复杂性。在本文中，我们提出了一种基于分段的特征，该特征是使用修改的深CNN-CRF培训的特征，这导致复杂性显着降低。我们根据错误指标评估了我们在NYUV2数据库上的算法。所提出的算法的误差性能略高于深度CNN-CRF，但明显优于其他现有算法。

著录项

来源
《Conference on Information and Communication Technology》|2017年|372p|共6页
会议地点
作者
Akhil Rana; Aditya Trivedi;
展开▼
作者单位

展开▼
会议组织
原文格式 PDF
正文语种
中图分类 TN91-53;
关键词
Image segmentation; Training; Prediction algorithms; Databases; Image color analysis; Information and communication technology; Computational modeling;

机译：图像分割;训练;预测算法;数据库;图像颜色分析;信息和通信技术;计算建模;

相似文献

外文文献
中文文献
专利

1. Temporally Consistent Depth Map Prediction Using Deep Convolutional Neural Network and Spatial-Temporal Conditional Random Field [J] . Xu-Ran Zhao, Xun Wang, Qi-Chao Chen 计算机科学技术学报（英文版） . 2017,第003期

机译：基于深度卷积神经网络和时空条件随机场的时间一致深度图预测
2. Learning Depth from Single Monocular Images Using Deep Convolutional Neural Fields [J] . Fayao Liu, Chunhua Shen, Guosheng Lin, IEEE Transactions on Pattern Analysis and Machine Intelligence . 2016,第10期

机译：使用深度卷积神经场从单眼图像中学习深度
3. Time-Independent Prediction of Burn Depth Using Deep Convolutional Neural Networks [J] . Cirillo Marco Domenico, Mirdell Robin, Sjoberg Folke, Journal of burn care & research: official publication of the American Burn Association . 2019,第6期

机译：使用深卷积神经网络的燃烧深度独立于烧伤深度预测
4. Depth “prediction” using modified deep convolutional neural field [C] . Akhil Rana, Aditya Trivedi 2017 Conference on Information and Communication Technology . 2017

机译：使用改进的深度卷积神经场进行深度“预测”
5. Convolutional neural network based age estimation from facial image and depth prediction from single image. [D] . Qiu, Jiayan. 2016

机译：基于卷积神经网络的基于面部图像的年龄估计和基于单个图像的深度预测。
6. AUCpreD: proteome-level protein disorder prediction by AUC-maximized deep convolutional neural fields [O] . Sheng Wang, Jianzhu Ma, Jinbo Xu -1

机译：AUCpreD：通过AUC最大化深度卷积神经场预测蛋白质组水平的蛋白质异常
7. Weakly supervised training of deep convolutional neural networks for overhead pedestrian localization in depth fields [O] . Corbetta, Alessandro, Menkovski, Vlado, Toschi, Federico 2017

机译：弱监督的深度卷积神经网络训练高架行人在深度领域的定位

Depth “prediction” using modified deep convolutional neural field

摘要

著录项

相似文献

相关主题

期刊订阅