Machine Learning Applications in Medical Image Analysis – In this paper, we present an open source algorithm for multispectral data augmentation. In particular, we provide an automatic technique for automatically augmenting images with different parameters. We apply this algorithm onto synthetic and real data. Our algorithm combines the information obtained from real images with an algorithm that computes the parameters of the data augmentation process. We use the multi-class matrix transform to estimate the transformation and learn a set of transformations for each object. We describe the application of our algorithm on image augmentation for medical image analysis and the use of multispectral data augmentation in image classification.
Visual object segmentation is challenging due to the large number of objects in the world and the large amount of data. Most methods focus on small sample sizes and few object segmentation features. In this paper, we propose a new image segmentation algorithm that uses a Convolutional Neural Network (CNN), a CNN architecture, to learn a set of local features to classify the object. In this way, CNN segmentation learned from input image is able to reduce the space of different object categories while improving classification accuracy. In addition, we propose a two step learning process: (1) Convolutional neural networks can be trained in an unsupervised way while training CNN image for segmentation. (2) Convolutional networks can be designed to make efficient use of training data when using ImageNet to segment object objects. We present a two-stage learning scheme for our algorithms for different CNN architectures and demonstrate the effectiveness of our algorithms.
Graph Deconvolution Methods for Improved Generative Modeling
Visual Tracking via Superpositional Matching
Machine Learning Applications in Medical Image Analysis
Concise and Accurate Approximate Reference Sets for Sequential Learning
Matching Strategies for Multi-Object Tracking with Variational AutoencodersVisual object segmentation is challenging due to the large number of objects in the world and the large amount of data. Most methods focus on small sample sizes and few object segmentation features. In this paper, we propose a new image segmentation algorithm that uses a Convolutional Neural Network (CNN), a CNN architecture, to learn a set of local features to classify the object. In this way, CNN segmentation learned from input image is able to reduce the space of different object categories while improving classification accuracy. In addition, we propose a two step learning process: (1) Convolutional neural networks can be trained in an unsupervised way while training CNN image for segmentation. (2) Convolutional networks can be designed to make efficient use of training data when using ImageNet to segment object objects. We present a two-stage learning scheme for our algorithms for different CNN architectures and demonstrate the effectiveness of our algorithms.