Convolutional feature learning and Hybrid CNN-HMM for scene number recognition

Guo, Qiang; Wang, Fenglei; Lei, Jun; Tu, Dan; Li, Guohui

doi:10.1016/j.neucom.2015.07.135

Cited by 55 publications

(20 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The output layer of the CNN is using the softmax function to compute the posterior probability P ( x i | o i ) of class x i given the input observations. It has been shown that by using Baye's rule, the emission probability can be computed by the scaled likelihood:

P false(o_{i} false| x_{i} false) = \frac{P false(x_{i} false| o_{i} false)}{P false(x_{i} false)},

where P ( x i ) is the prior probability of class x at state i , and it is computed by counting the number of state class in the training examples.

P false(x_{i} false| o_{i} false) = y_{x} = \frac{\exp false(z_{x} false)}{\sum_{x^{'}} \exp false(z_{x^{'}} false)},

where z x is the output neurons from the previous layer multiplied by the weights.…”

Section: Proposed Methodsmentioning

confidence: 99%

Coupled s‐excess HMM for vessel border tracking and segmentation

Essa

Jones

Xie

2019

Numer Methods Biomed Eng

View full text Add to dashboard Cite

In this paper, we present a novel image segmentation technique, based on hidden Markov model (HMM), which we then apply to simultaneously segment interior and exterior walls of fluorescent confocal images of lymphatic vessels. Our proposed method achieves this by tracking hidden states, which are used to indicate the locations of both the inner and outer wall borders throughout the sequence of images. We parameterize these vessel borders using radial basis functions (RBFs), thus enabling us to minimize the number of points we need to track as we progress through multiple layers and therefore reduce computational complexity. Information about each border is detected using patch‐wise convolutional neural networks (CNN). We use the softmax function to infer the emission probability and use a proposed new training algorithm based on s‐excess optimization to learn the transition probability. We also introduce a new optimization method to determine the optimum sequence of the hidden states. Thus, we transform the segmentation problem into one that minimizes an s‐excess graph cut, where each hidden state is represented as a graph node and the weight of these nodes are defined by their emission probabilities. The transition probabilities are used to define relationships between neighboring nodes in the constructed graph. We compare our proposed method to the Viterbi and Baum–Welch algorithms. Both qualitative and quantitative analysis show superior performance of the proposed methods.

show abstract

P false(o_{i} false| x_{i} false) = \frac{P false(x_{i} false| o_{i} false)}{P false(x_{i} false)},

where P ( x i ) is the prior probability of class x at state i , and it is computed by counting the number of state class in the training examples.

P false(x_{i} false| o_{i} false) = y_{x} = \frac{\exp false(z_{x} false)}{\sum_{x^{'}} \exp false(z_{x^{'}} false)},

where z x is the output neurons from the previous layer multiplied by the weights.…”

Section: Proposed Methodsmentioning

confidence: 99%

Coupled s‐excess HMM for vessel border tracking and segmentation

Essa

Jones

Xie

2019

Numer Methods Biomed Eng

View full text Add to dashboard Cite

show abstract

“…The output holds scores of the classes [8]- [10]. For creating CNN classifiers, a few distinct types of layers are commonly used [2], [3], [8], [9]. They are convolutional layers (ConvL), rectified linear unit layer (ReLU), average pooling layer (AvPL), max pooling layer (MaxPL), fully connected layer (FCL), softmax layer (SML) and dropout layer (DOL) [1]- [3], [8], [10], [11].…”

Section: An Open Problem Of Setting Hyperparametersmentioning

confidence: 99%

Smooth Non-increasing Square Spatial Extents of Filters in Convolutional Layers of CNNs for Image Classification Problems

Romanuke

2018

Applied Computer Systems

View full text Add to dashboard Cite

The present paper considers an open problem of setting hyperparameters for convolutional neural networks aimed at image classification. Since selecting filter spatial extents for convolutional layers is a topical problem, it is approximately solved by accumulating statistics of the neural network performance. The network architecture is taken on the basis of the MNIST database experience. The eight-layered architecture having four convolutional layers is nearly best suitable for classifying small and medium size images. Image databases are formed of grayscale images whose size range is 28 × 28 to 64 × 64 by step 2. Except for the filter spatial extents, the rest of those eight layer hyperparameters are unalterable, and they are chosen scrupulously based on rules of thumb. A sequence of possible filter spatial extents is generated for each size. Then sets of four filter spatial extents producing the best performance are extracted. The rule of this extraction that allows selecting the best filter spatial extents is formalized with two conditions. Mainly, difference between maximal and minimal extents must be as minimal as possible. No unit filter spatial extent is recommended. The secondary condition is that the filter spatial extents should constitute a non-increasing set. Validation on MNIST and CIFAR- 10 databases justifies such a solution, which can be extended for building convolutional neural network classifiers of colour and larger images.

show abstract

“…Segmenting and recognizing the characters go hand in hand. With a goal to amalgamate segmentation and recognition, we will follow the proposed theory of hybrid CNN-HMM [18] where the model undergoes sliding window mechanism to extract a series of frames. Feature descriptors like SIFT, HOG and LBP are widely used in vision problems.…”

Section: International Journal Of Computer Applications (0975 -8887) mentioning

confidence: 99%

Efficient Object Recognition using Convolution Neural Networks Theorem

Thakral¹,

Shekhar²,

Victor³

2017

IJCA

View full text Add to dashboard Cite

Object recognition is the process of identification of an object in an image. There exist various algorithms for the same. Appearance based algorithms have demonstrated good efficiency, however, their performance gets affected adversely in the presence of clutter or when background changes are affected. We hope to overcome this issue by using Convolution Neural Network (CNN) Theorem. The approach is shape based and has been proven to work well under broad range of circumstances: varied lighting conditions, affine transformations, etc. It involves tiling, which is the phenomenon of the use of multiple layers of neurons to process small portions of the image, which are then used to obtain better representations of the image. This allows CNN to be translation-tolerant. The neural elements learn to recognize objects about which they have no previous information, this "learning" mechanism is affected by the fact that representations of the image are learned by the inner layers of the deep architectures of neurons. Unlike RBM and Auto-encoder, which are capable of learning only single global weight matrix layers, the CNN theorem makes use of shared weight in convolution layers, which means that the same filter (weight bank) is used for each pixel in the layer, which reduces the memory footprint and improves performance.

show abstract

Convolutional feature learning and Hybrid CNN-HMM for scene number recognition

Cited by 55 publications

References 41 publications

Coupled s‐excess HMM for vessel border tracking and segmentation

Coupled s‐excess HMM for vessel border tracking and segmentation

Smooth Non-increasing Square Spatial Extents of Filters in Convolutional Layers of CNNs for Image Classification Problems

Efficient Object Recognition using Convolution Neural Networks Theorem

Contact Info

Product

Resources

About