Stack Exchange Network

Stack Exchange network consists of 174 Q&A communities including Stack Overflow, the largest, most trusted online community for developers to learn, share their knowledge, and build their careers.

Visit Stack Exchange

Convolutional Neural Networks are a type of neural network in which only subsets of possible connections between layers exist to create overlapping regions. They are commonly used for visual tasks.

-1
votes
1answer
18 views

Machine learning books covering neural networks / cnn / GAN [duplicate]

I'm not an expert in machine learning. Is there any textbook (with a decent amount of mathematical rigor) that cover the subjects neural network / convolutional neural network / GAN network? I've the ...
2
votes
1answer
18 views

What would happen if CNN reused same kernel weights for each channel?

In a CNN each output location of a feature map is given by the kernel over the previous layer's feature maps. If the receptive field is say 5x5 with 3 channels then there are 5x5x3 weights that are ...
0
votes
0answers
9 views

Is softmax a fully connected layer? [on hold]

I know that in all CNN networks there usually a softmax layer at the end of the network. My question is the that softmax layer a fully connected layer or not?
0
votes
0answers
19 views

How to prevent overfitting with CNN (keras) [on hold]

I am training a Convolutional Neural Network to classify text and getting a extremely hight training accuracy per epoch, but getting a low test accuracy. I presume my model is overfitting. Does anyone ...
2
votes
1answer
13 views

Clustering / Grouping on image's pixels

I have an image, and im building a model to recognize a pattern in that image and classify it. There is however a lot of noise in the rest of the image, but the actual pattern to classify will always ...
0
votes
0answers
10 views

Different random weight initilization leading to different performances

I'm training a 3D U-Net on an EM dataset of a brain. The objective is to segment neurons in it. During the experiments, I've noticed, different random initialization of the network leads to different ...
0
votes
1answer
15 views

Normalization of inputs in convolutional neural networks

I read that using convolutional neural networks, or any neural networks (?), that the input/features should be normalized. The normalization is typically done for each feature $$x_i \in X \ \ \forall ...
0
votes
0answers
10 views

Batch Normalization and increasing batch size reducing the performance

I'm training a 3D U-Net on an EM dataset of a brain. The objective is to segment neurons in it. During the experiments, I've noticed, increasing batch size, adding batch normalization layers (Conv ->...
3
votes
2answers
25 views

Encoder Decoder networks with varying image sizes

Encoder Decoder Network - Computerphile : At the very beginning of this video, Michael Pound goes on to say: So it (encoder decoder network) makes no assumptions about the size of the input the ...
0
votes
2answers
26 views

convolutional autoencoder on an odd size image

I am trying to apply convolutional autoencdeor on a odd size image. Below is the code: ...
0
votes
0answers
6 views

Training accuracy of the model keeps decreasing with each epoch even though loss is decreasing

I am training a model to recognise characters from images with 8 conv layers. I am having problem that the train accuracy decreases by large value in each epoch, the first few have like .80-.90, even ...
0
votes
0answers
14 views

U-Net image size for training

I have a small question regarding the size of images used for training the U-Net. I have thus far been able to train a U-Net reasonably well using 656x656 images and now wanted to use sections of ...
0
votes
0answers
7 views

Dropout and feature visualisation

If we use drop out in our CNN, this will lead to features being more displeased and less concentrated on specific neuroma. Won't this make feature visualisation very difficult since we can't locate ...
0
votes
0answers
12 views

Select the right architecture for a deep learning binary classifier on DNA data

I'm working on a deep learning binary classifier. The train dataset properties are: Input matrix : 26500 individuals x 18 000 genes. Genes (SNP) are encoded as follow 0 for homozygous major, 1 for ...
1
vote
0answers
23 views

Using a neural network for a regression problem, where the model to be learned suffers from awgn

I have currently a neural network to learn a (relatively non complex) system model (vector regression). Its problem is that the outputs of the system suffer from arbitry additional white gaussian ...
0
votes
1answer
14 views

Generalization of different play-field sizes

I'm currently working on a Deep Reinforcement Learning model for the game "connect 4". Before I started I read some rules and facts about "connect 4". Thats when I noticed after years of playing it, ...
1
vote
0answers
30 views

Recognition the same object from different views

I have 33 classes (33 different objects). I need to recognize the object from any view of the object. Like a packet of potato chips, the packet has different appearance from different view (as shown ...
1
vote
0answers
30 views

Is YOLO a good algorithm for defect detection on images?

I wish to train an algorithm to detect defects on images of labels. These may be such things as scratches, tears and voids. I would like to try to train a YOLO algorithm to do this, but it is very ...
0
votes
0answers
14 views

Training SSD like network for Number and Alphabet detection

I tried to read number plate using the model trained with SSD like network. My network has only five layers. I trained with 60,000 images for digits 0-9 and alphabets A-Z. So one class has 2,000 ...
0
votes
0answers
10 views

CNN: Modifying VGG16 Architecture

I'm currently trying to modify the VGG16 network architecture so that it's able to accept 400x400 px images. Based on literature that I've read, the way to do it would be to covert the fully ...
0
votes
0answers
20 views

Feature recognition with partial images on CNN

If I train a CNN to learn to recognise features on complete images, would that same network be able to recognise features on a partial image of the same type? The motivation for asking this question ...
1
vote
0answers
24 views

CNN - Is this a Toeplitz Matrix?

I have been reading through Chapter 9 of www.deeplearningbbook.org, where convolutional networks are being described. The following image represents the output of a 2D convolution, without kernel ...
1
vote
0answers
20 views

Why is this convolution equation easier to implement than it's commutative counterpart?

The convolution is an operation on two functions of a real- valued argument and is typically denoted with an asterisk: s(t) = (x ∗ w)(t) It is a special kind of ...
-1
votes
0answers
5 views

Problem training a (variational) autoencoder on a handwritten signature database [duplicate]

I'm trying to train both a normal autoencoder and a variational autoencoder on a hand written signature database, but it seems like the system is not learning anything, since the reconstructed images ...
1
vote
0answers
58 views

Is it true that normalizing the output of a ReLu feedforward Neural Network that its Rademacher Complexity becomes a constant?

I was trying to understand what happened with the Rademacher Complexity: $$ R_S(F) = \frac{1}{m} \mathbb E_{\sigma} [\sup_{f \in F}\sum^m_{i=1} \sigma_i f(z_i)] $$ or $$ R_{P,m} = \mathbb E_{s \...
1
vote
0answers
8 views

Counting number of items in a box from a MP4 video using computer vision and R

I have an MP4 video in which a person keeps putting some household items into the box(container type). Items consist chips packets, colddrink bottles, mugs etc. Sometimes even person takes out some ...
0
votes
0answers
17 views

I am trying to make a classifier which classifes image into 5 categories. But it is giving only one category for all images

I am using convolutional neural network to classify the image into five different categories. In the final output layer(a Dense Layer) i am using 'softmax' as the activation function. While compiling ...
0
votes
0answers
17 views

Autoencoder with small dataset - simple images

I wanted to ask a simple question regarding autoencoders to parse for tips and possible advice before diving into this path. I have a small dataset ~50-100 heatmap images that are relatively simple. ...
0
votes
0answers
19 views

Many binary or a single int input layer (Deep Learning)

I am building a data generator with 5 distinct features that are mapped on a 2D grid with 300 pixels (each feature uses many x, y coordinate pairs) and I want to use a CNN type architecture to do some ...
1
vote
1answer
30 views

Residual learning approach + manifold

From: Wikipedia on Residual learning in ANNs "The intuition on why this works is that the neural network collapses into fewer layers in the initial phase, which makes it easier to learn, and thus ...
0
votes
0answers
16 views

Why ResNet use large kernel size convolutions at first layers?

What is motivation of using large kernel size convolutions at first layers in ResNet? i.e. in subsequent layers it use (3,3) convolutions like VGG.
0
votes
1answer
32 views

Why does distorting images improve training on a neural network?

I cannot understand why distorting an image, e.g flipping it, increasing the gamma intensity would somehow increase the accuracy on neural network. Within my situation, I am Using a CNN to detect if ...
1
vote
0answers
27 views

What are the best practices to normalize inputs before training a CNN?

I have a dynamic range of 256x256 matrices. I want to have a CNN based binary classifier. The matrices are images with a very wide range of intensities (10 orders of magnitude). I am afraid to use ...
0
votes
1answer
31 views

What are the ways to calculate the error rate of a deep Convolutional Neural Network, when the network produces different results using the same data?

I am new to the object recognition community. Here I am asking about the broadly accepted ways to calculate the error rate of a deep CNN when the network produces different results using the same data....
0
votes
0answers
32 views

AlexNet Paper dimensions does not match?

I am following the paper where AlexNet was introduced, and the dimensions they report just don't match with the figure they attached. The output of the first conv layer (which is 96 11x11x3 ...
0
votes
1answer
24 views

How can we interpret the learning curve including loss for training and test in a deep learning model?

I am working on 3D medical image segmentation area. It may take 2-3 days to finish one round of training. How can I interpret the learning curve if over-fitting is happening or not? It happens to me ...
0
votes
0answers
8 views

Contradiction between accuracy obtained from my pretrained conv base network and pretrained conv base network in Deep Learning with Python

I trained a pretrained convnet model on the cats and dogs dataset and the following are the accuracies obtained: Freezed Conv Base ~ 90% Unfreezed Conv Base ~ 96% However this is in contradiction to ...
0
votes
0answers
10 views

My evaluation accuracy curve lays down above the training set accuracy one. Is it normal?

I'm try to build a text classifier using a CNN with word embedding with Keras and Tensorflow. Graph from tensorboard Here is a snippet of the code that shows the model construction: ...
0
votes
0answers
17 views

Have CNNs matched radiologists in predicting breast cancer from mammograms yet?

I'm at a talk right now about machine learning in medicine. One of the slides showed a CNN (Convolutional Neural Network) rivaling radiologists in accuracy of mammogram readings. It predicted things ...
0
votes
0answers
86 views

how to visualize feature map of resnet with tensorflow?

I have trained a resnet50 model for classification. Sometime it predicts wrong for some image. So I want to visualization the response map to show what is the feature details it learned. But I did not ...
1
vote
0answers
28 views

What does the ADAM algorithm actually guarantee?

Can someone provide an intuitive explanation as to what the propositions of the ADAM algorithm actually guarantees? and whether they are strong guarantees? https://arxiv.org/pdf/1412.6980.pdf There ...
1
vote
1answer
34 views

Can Neural Network take multidimensional inputs?

I recently learned RNN and find that a common feature of it and CNN is that they use either a LSTM cell or Convolution to process a single multidimensional input (like images and word embeddings which ...
1
vote
1answer
50 views

Feature map of a convolutional layer [duplicate]

Say there is a convolutional layer that has 8 filters of size 3x3 and takes as input 28x28x1 images. Then the output size will be 26x26x8. So how are there 8 26x26 feature maps for that layer or ...
2
votes
1answer
41 views

Constant Accuracy with decreasing loss

I am fairly new to Cross Validated section so I apologize if my question structure is incorrect. I am currently working on Fully Convolutional Networks for Semantic Segmentation. I am first trying ...
1
vote
1answer
21 views

What is the purpose or benefits of fully connected layer at the middle of Convolutional Network?

Is there any benefits to have FC layer at the middle of CNN network? For example, in this network, FC7 has kernel size is 1. What is the benefits of using kernel size 1 in this use? Those inception ...
0
votes
1answer
28 views

Mobilenet Original Paper Architecture vs Keras Implementation

1. Question: Why do original paper mobilenet architecture and keras implementation differ? Keras implementation of mobilenet's last 5 layers after AVG Pool layer: ...
0
votes
1answer
19 views

For EEG analysis, why is it more efficient to use the raw data than images of the data?

Data Science publications for EEG analysis like EEGNet: A Compact Convolutional Network for EEG-based Brain-Computer Interfaces use the raw EEG data (see: github vlawhern/arl-eegmodels ) rather than ...
0
votes
1answer
44 views

RGB images as input to CNN

Considering a 32*32*3 RGB image, would there be filters/kernels for each color channel? I haven't found examples explaining how CNN works for RGB images and whether each filter is a 3D. If I decide to ...
0
votes
0answers
25 views

FLOPS for CNN networks

When I have a new CNN network, I checked how many GFLOPS are needed for a particular size of image for my application. Say, one particular network with 240 x 240 image size has 1.2GFLOPS, am I fair ...
1
vote
1answer
49 views

Is a SVM (+Boost) faster than a NN to train with similar accuracy?

I might possibly misunderstand something here. So please do tell me if I do. At the moment, I am doing some research regarding the use of machine learning to detect a certain object. Currently I am ...