Questions tagged [cnn]

Convolutional Neural Networks (CNN, also called ConvNets) are a tool used for classification tasks and image recognition. The name giving first step is the extraction of features from the input data.

Filter by
Sorted by
Tagged with
1
vote
1answer
15 views

Finding the appropriate CNN Model Architecture and Parameters

I am currently creating a CNN model that classifies whether the font is Arial, Verdana, ...
0
votes
0answers
10 views

Multimodal vs recurrent network for video-based FER

I am wondering what types of architectures are best to explore for doing video-based facial emotion recognition. Some application feature using an architecture spatial-temporal architecture of 2 CNNs, ...
0
votes
0answers
17 views

Getting big losses and little accuracy on image classification model with cnn

i am currently working on image classification of artworks from this site https://www.kaggle.com/ikarus777/best-artworks-of-all-time and following the tutorial from this site https://...
1
vote
1answer
18 views

How are the channels handled in CNN? Is it independently processed or fused?

Let's assume that we are talking about 2D convolutions applied on images. In a grayscale image, the data is a matrix of dimensions $w \times h$, where $w$ is the width of the image and $h$ is its ...
2
votes
1answer
28 views

EfficientNet: Compound scaling method intuition

I was reading the paper EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks and couldn't get my head around this sentence: Intuitively, the compound scaling method makes sense ...
-2
votes
1answer
29 views

What is the reason to use a RNN over a CNN in text classification task? [closed]

When should i switch to using an RNN (LSTM, GRU) over a simple CNN for classifying web text articles on pre-specified taxonomy? My input is 100K of news feed articles mostly pre-labeled in 15 ...
-1
votes
0answers
23 views

Car classification [closed]

i want to classify cars into make and model (year would be good too but i need too much data for that) for example. BMW X4, BMW X5, BMW 540, Audi tt, Audi s6, Audi s8, ford..., opel..., Volkswagen... ...
2
votes
1answer
20 views

Keras BatchNormalization axis

I use spectrogram as input to a Convolutional Neural Network I have created with tensorflow.keras in Python. Its shape is (time, frequency, 1). The input's shape of the CNN is (None, time, ...
0
votes
0answers
19 views

Fusing batch normalization with deconvolution in neural networks

I am trying to raise the performance of my convolutional neural network and for that reason I am trying to implement batch normalization fusing. Things are fine when I use fuse with convolution layer,...
0
votes
0answers
21 views

How to recognize overlapping digits?

I've got a set of images with overlapping digits which need to be recognized. The task seems a good fit for neural networks but the issue is that they are used to have inputs as single digits but in ...
0
votes
0answers
9 views

Is it possible to fit GMM so that mean lies outside of image bounding box?

I m currently reading https://arxiv.org/pdf/1411.3159.pdf They take gradient maps and fit Gaussian Mixture Model. Then they evaluate how often is the activation center inside of the image bounding box....
1
vote
0answers
14 views

How to represent and work with the feature matrix for graph convolutional network (GCN) if the number of features for each node is different?

I have a question regarding features representation for graph convolutional neural network. For my case, all nodes have a different number of features, and for now, I don't really understand how ...
0
votes
1answer
39 views

Binary image classifier always predicting one class

I am trying to design a model for binary image classification, this is my first classifier and I am following an online tutorial but the model always predicts class 0 My dataset contains 3620 and ...
0
votes
0answers
24 views

Keras : How to Connect CNN ResNet50 with svm/random forest classifier?

I want to classify multiclass (10 classes) images with random forest and SVM classifier, that is, make a hybrid model with ResNet+SVM , ResNet+random forest. My ResNet code is below: ...
0
votes
0answers
6 views

Patch wise training vs Full Convolutional Training in semantic segmentation

As mentioned in the title, what are those 2 methods? I already checked this question: Patchwise and Full training, (and the mentioned paper) but i can't really understand the meaning and the process ...
0
votes
0answers
11 views

Can one hardcode convolutional filters to detect characters in a CNN?

In Pytorch, you can hardcode your filters to be whatever you like. At the moment, I'm doing text detection and I need to identify the location of a certain information. This information always ...
0
votes
0answers
11 views

Batch Normalization when CNN with only 2 ConvLayer?

I wonder if it is a problem to use BatchNormalization when there are only 2 convolutional layers in a CNN. Can this have adverse effects on classification performance? Now I don't mean the training ...
0
votes
2answers
36 views

How to understand the weights and biases for beginners?

I am newbie to deep learning, I was building my first model using MNIST dataset, I understood the full model, but one thing is a bit confusing to me. How can we get the weights and bias? Is it that, ...
0
votes
0answers
21 views

confusion matrix results in CNN keras

I am writing a code to classify images from two classes, dogs and cats. I wrote the below code, but always all the dogs images are classified as cats as shown in the confusion matrix. Am I missing ...
1
vote
0answers
11 views

Efficient implementation of seperable convolution in tensorflow

It seems like the native implementation of separable convolution in tensorflow is not efficient. https://github.com/tensorflow/tensorflow/issues/12940 Is anyone aware how can we get an efficient ...
0
votes
0answers
13 views

How to identify and extract text from table in a image ? Are there any machine learning model avaible for extracting text in a table?

How to identify and extract text from table in a image ? sample image shown below Are there any machine learning model available for identifying table and extracting text in a table ? i tried ...
1
vote
0answers
23 views

Keras Encoder - Decoder Model

Im training a model of 400 samples . The dataset contains 400 images of faces as input (X) and also 400 faces with glasses as output (Y) . im training the model by code below : ...
2
votes
2answers
43 views

What's the purpose of padding with Maxpooling?

As mentioned in the question, i've noticed that sometimes there are pooling layers with padding. More specifically, I found this Keras tutorial, where there's a net which contains ...
1
vote
1answer
12 views

How does ResNet bottleneck architecture's input size is possible to change from 56x56x64 to 56x56x356?

In ResNet papaer, First residual block's input size is 56x56x64 caused by 7x7x64 filter in first layer. But, in the paper, they showed residual block that has 56x56x256 input size. How does it is ...
0
votes
0answers
15 views

Definition of “repooling” in YOLACT paper

In the latest YOLACT paper found HERE I see the term "repooling" in this part and others: Moreover, we obtain this result after training on only one GPU. We accomplish this by breaking instance ...
0
votes
1answer
23 views

How can I train my CNN to learn (numerically) smaller values better?

I'm using a CNN to model a problem that involves precise numerical values from a physical simulation. After months of design/redesign and optimization, I've noticed that the majority of the "error" ...
0
votes
0answers
12 views

Facing memory error while converting Dicom data into array

I am building a disease classifier using Dicom scans for many patients. Different patient's scans have different number of slices. For example: Patient1's scan : 100 slices Patient2's scan : 500 ...
4
votes
1answer
49 views

Keyword localization in audio file

I want to build a model that can localize occurrences of a particular word in an audio file. For example, I want to find the word "pizza" in a ~5min recording. The program should return an array with <...
1
vote
1answer
18 views

What is the layer above/below in a NN?

In the lecture notes of CS231n, it says (emphasis mine) ... There are three major sources of memory to keep track of: From the intermediate volume sizes: These are the raw number of ...
1
vote
0answers
22 views

How are the Convolution kernels learned?

As I went through the basics of machine learning, I failed to understand how do the Convolutional layers in a CNN learn the convolution kernels. After looking at first few tutorials, I thought the ...
0
votes
0answers
10 views

How to do template matching without opencv?

How to do template matching without opencv ? I have a order invoice of documents belonging to Amazon ,ebay,flipkart,snapdeal and i want to extract few information from the order invoice . Since the ...
2
votes
0answers
19 views

What doese 'v' mean in GoogLeNet?

In GoogLeNet (this link), there is 'v' notation in Figure3 like '1X1+1(v)'. I don't know the meaning of 'v'. Also, I understood 's' as stride. But, I don't know the reason why plus operation is used ...
1
vote
0answers
13 views

how to update a face recongition ( e.g facenet)

I'm searching for a face recognition that detects new faces not just faces getting from the datasets, for example, facenet is a project that can detect and recognize a face from labels. I would like ...
1
vote
0answers
14 views

I don't know local spatial correlation meaning

In GoogLeNet papaer, 'local sparse correlation' is mentioned. But, I don't know the meaning. Also, I don't understand that statement " In the lower layers (the ones close to the input) correlated ...
1
vote
0answers
38 views

Keypoint detection from an image using a neural network

I am trying to design and train a neural network, which would be able to give me coordinates of certain key points in the image. Dataset I've got a dataset containing 1800 images similar to these: ...
0
votes
1answer
10 views

How does two convolution layers make computation quadratic increase in GoogLeNet?

In GoogLeNet papaer, "For example, in a deep vision network, if two convolutional layers are chained, any uniform increase in the number of their filters results in a quadratic increase of computation....
1
vote
0answers
25 views

What is the difference between layer degradation and overfitting?

I've read the ResNet paper (https://arxiv.org/pdf/1512.03385.pdf) They determined the layer degradation as that model with less layers learn quicker than with more. It can be visiable on plots below: ...
0
votes
0answers
12 views

CNN: What's the relationship between point clouds and features derived from point clouds?

What's the relationship between point clouds and features derived from point clouds? Particularly in CNN prediction. Particularly, I have point clouds about which I can imagine features that are ...
1
vote
0answers
26 views

setting class weights for imbalanced dataset, how using EarlyStopping?

I want to train a CNN with Early Stopping (Keras). The data set is imbalanced, so I have set class_weights to 'balanced' like follows: ...
2
votes
1answer
30 views

Image Super-resolution Connecting Subimages

I'm working with image super-resolution on terrain height data. Currently, I'm cutting the input data into smaller pieces (20 x 20 rather than 10800 x 10800). After the upscale 20x20 -> 40x40, ...
1
vote
1answer
32 views

How to use CNN to deal with a 2D regression problem?

I have seven measurements (Obs1-7), each measurement has the dimension of [x,y,t] where x and y are coordinates and t is time. Now I want to build a model that uses the first 6 measurements to predict ...
2
votes
0answers
13 views

How to create anchor-positive and anchor-negative pair for feature X in a signature data set for training Siamese network

How to create anchor-positive and anchor-negative pair for feature X in a signature data set for training Siamese network? Im have a cedar signature data set with 55 peoples signatures(classes) with ...
3
votes
1answer
67 views

Will this MAPE implementation work for multidimensional output?

I'm currently working on a CNN problem where the output is a 60x59 array of numerical values. I want to verify if the mean absolute percentage error (MAPE) function I'm employing will properly ...
4
votes
2answers
195 views

Skin Detection Classifier

I have a small data set containing around 80 images for people, and the corresponding ground truth for skin regions. I want to train a classifier to be able to detect a skin, and use it later on my ...
1
vote
1answer
22 views

How to process Dicom Images for CNN?

I am building a disease classifier. I have Dicom scans for many patients. The scans have different slice thickness, and different scans have different number of slices. However, the slice thickness ...
1
vote
0answers
14 views

How to build Explanatory Graph for Convolutional Neural Network?

I m reading very interesting paper (https://arxiv.org/pdf/1812.07997.pdf) that aims to interpret convolutional neural network using graph. The general idea is when there are co-related parts in layers ...
1
vote
0answers
82 views

Non differentiable loss function

I have a loss function that minimizes the error according to what I want the neural network to do. The problem is, that it is a nondifferentiable function. How can I handle this? the loss function: $(...
2
votes
0answers
27 views

backpropagation between fully connected layer and convolution layer?

This is a simple example of a network consisting of two convolutional layers and one fully connected layer. ...
3
votes
1answer
63 views

Need of maxpooling layer in CNN and confusion regarding output size & number of parameters

In my CNN architecture for binary classification, I have 2 convolutional layers, 2 maxpooling layers, 2 batchnormalization operations, 1 RELu and 1 fullyconnected layer. Case1: When the number of ...
1
vote
0answers
14 views

An CNN seems like capturing specific range of input data. (Image Segmentation)

I'm trying to build a model to segment brain tumors. I trained a model, and the validation dice coefficient is disappointing(0.6). When i saw the predicted images with the ground truths, it seems ...

1 2 3 4 5 18