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
0answers
10 views

what are the ways to gain good accuracy in deep learning competition apart from changing hidden layer especially in case of image data set?

I am newbie in field of deep learning but i have experience in machine learning.like in machine learning competition we do plotting apply some statistics to gain understanding of data and by that we ...
1
vote
0answers
10 views

GoogleNet Loss Skyrockets

I am training GoogleNet on the Stanford cars data set. It's 8000 training images of cars with labels (2004 Toyota Camry). I made minimal changes to the network. I just changed the loss outputs to ...
0
votes
1answer
8 views

What are different methods to find the slow decrease in training/validation loss

I am training YOLO network consisting of resnet50 architecture.This problem is to find different text labels on the image and predict bounding boxes During training, I am seeing very less change in ...
1
vote
0answers
18 views

BatchNorm after ReLU

I am currently experimenting with different settings for a U-Net (https://lmb.informatik.uni-freiburg.de/people/ronneber/u-net/) based image segmentation and I was unable to find out if it makes any ...
-1
votes
0answers
16 views

Help: training accuracy too low

[Using Keras/tensorflow] I'm trying to train a model suggested in this paper. I've set the weights of two convolutional layers as gabor filters. When I train the model, I'm getting the per wpoch ...
2
votes
0answers
15 views

Why is it difficult to learn a single kernel that performs well at all positions in the convolutional feature map?

I am reading Deep Learning book by Ian Goodfellow, in which they wrote (in chapter 9, section 9.5) that: " ... MATLAB refers to this as full convolution, in which enough zeros are added for every ...
0
votes
1answer
24 views

Fundamental questions on CNN and MLP in general [closed]

I have read a number of tutorials and online lectures (link1: https://ujjwalkarn.me/2016/08/11/intuitive-explanation-convnets/) and link2: the webtutorial: http://cs231n.github.io/convolutional-...
4
votes
1answer
26 views

Why do CNNs conclude with FC layers?

From my understanding, CNNs consist of two parts. The first part (conv/pool layers) which does the feature extraction and the second part (fc layers) which does the classification from the features. ...
0
votes
1answer
15 views

Why does cross entropy loss for validation dataset deteriorate far more than validation accuracy when a CNN is overfitting?

I have noticed that the cross-entropy loss for validation dataset deteriorates after a certain number of epochs when training CNN's or MLP's. This is, of course, the sign that the network is ...
0
votes
0answers
21 views

When training a CNN, the validation loss and validation accuracy never changes. Is this a problem with the dataset?

When training a CNN, the validation loss and validation accuracy never changes. Is this a problem with the dataset? No matter what architecture or number of layers or set of parameters, the val train ...
1
vote
1answer
27 views

U-Net convolutional neural network

I am currently trying to understand how exactly the U-Net (https://lmb.informatik.uni-freiburg.de/people/ronneber/u-net/) works and so far failed to understand some key points, which are the following:...
1
vote
0answers
14 views

When do we use an even size kernel in convolutional neural network and why?

Recently, I've been seeing more and more code/paper using even size kernels in ConvNets, which is quite counter-intuitive to me. I wish someone could shed some light on the reasoning behind it: When ...
0
votes
0answers
21 views

How to build CNN to detect open defects? [closed]

I'm having enough dataset to train a CNN from scratch, but as I'm new to deep learning I'm confused that how I configure a CNN to detect open defects in image, because in dataset there are images with ...
1
vote
1answer
18 views

Why we don't normalize the images?

I was watching the video from this stanford course on convolutional neural nets where the professor says (at 28:59) 'we do zero-mean the pixel values in image but we do not normalize the pixel values ...
0
votes
1answer
20 views

Does not being able to overfit a single training sample mean that the neural network architecure or implementation is wrong? [duplicate]

Is the following hypothesis true ? If a simple neural network cannot overfit a single training sample, there is something wrong with its architecture or its implementation. To give you more ...
1
vote
0answers
5 views

Implementing a threshold for multiclass classification with softmax activation

We have a CNN where images are classified into between 7 and 30 classes, depending on the training set. The final output is via SoftMax activation, and thus all the probabilities add to one. I notice ...
2
votes
0answers
25 views

Developing algortihm/model to identify thin linear features in aerial imagery [closed]

I am exploring the possibility of identifying fencelines from NAIP aerial imagery (GSD = 0.6m). I have tried some basic processing in OpenCV using canny edge detection that was detailed in a question ...
1
vote
1answer
22 views

Connection between filters and feature map in CNN

I am learning CNN with TensorFlow and Python. I do not understand the connection between layer $\ell$ and layer $\ell+1$. For example, for the input image and the first layer, it is easy as there is ...
0
votes
0answers
17 views

What is the accepted definition of a Shallow Convolutional Neural Network? [duplicate]

While surfing for an acceptable definition, the general consensus was that a network with 1 Convolution Layer is called a Shallow CNN, but with the advent of so many deep Nets, has this definition ...
2
votes
1answer
48 views

Intuitive or quantitative explanation of why we care about mean average precision (mAP) for CNN classifiers?

Consider CNN classifiers applied to some image classification tasks: to fix ideas, let's consider the ImageNet Challenge, where each image belongs to 1 of 1000 nonoverlapping classes, even though the ...
0
votes
1answer
17 views

Optimum Weight update in CNN training [closed]

I have two networks. D->C1,R1->P1->C2,R2->loss and D->C1,R1,C2,R2->P1->C3,R3->loss. D is data, C for Conv, R for Relu and P for Pool. I trained first network from scratch and since two networks have ...
0
votes
0answers
26 views

Triplet loss - what threshold to use to detect similarity between two embeddings?

I have trained my triplet loss model using FaceNet's architecture. I used 11k hands dataset. Now I want to see how well my model performed, so I feed it 2 images of the same class and get back their ...
2
votes
1answer
104 views

Is convolution neural network (CNN) a special case of multilayer perceptron (MLP)? And why not use MLP for everything?

If convolution can be expressed with matrix multiplication (example) Can we say convolution neural network (CNN) is a special case of multilayer perceptron (MLP)? If yes, why people do not use a big ...
0
votes
1answer
36 views

Generating sequences of musical chords

I'd like to create a model capable of emulating music that has been presented to it. The model ought to be specifically designed for that purpose, not just another generic, stacked LSTM. For the ...
0
votes
0answers
47 views

Modeling a CNN to identify if image “is” or “isn't” something

I'm trying to build a CNN to play a game online. This game to be precise: https://www.gameeapp.com/game-bot/ibBTDViUP I've collected images and labels for each image. These labels tell the network ...
4
votes
3answers
127 views

What are the current state-of-the-art convolutional neural networks?

I'm interested in understanding which neural network architecture is currently "the best" with respect to standard image classification tasks such as MNIST, STLN-10 and CIFAR. This is challenging ...
0
votes
0answers
19 views

LSTM Training with High Score Variace

I am training a NN as conv -> conv -> dense -> lstm -> softmax, and got the scores/iteration as the figure shows. The dataset has been normalized with ...
0
votes
0answers
11 views

Computational and Memory complexity of object detection vs face recognition (for inference only)

I'm thinking of implementing a vision-based deep learning inference algorithm on an embedded device (CPU only). Would like to narrow down some feasible / practical applications. Which has lower ...
1
vote
1answer
19 views

Can a Fully Connected layer transform a 4D tensor to a 3D tensor by itself?

Recently, I was researching some topics in biometrics and I stumbled upon this paper. They have a table there (Table 1) in which they state that they used a modified CNN from this paper (Table 9). In ...
0
votes
1answer
24 views

Why we use activation function after convolution layer in Convolution Neural Network?

I'm new to machine learning and one of the things that I don't understand about Convolution neural networks, is that why we perform activation after convolution layer.
1
vote
1answer
36 views

What is the adventage of using Reinforcement learning in designing CNN?

I am looking at this paper Designing Neural Network Architectures using Reinforcement Learning. The paper discussed how to find the best network using ...
3
votes
2answers
499 views

In CNN, do we have learn kernel values at every convolution layer?

I'm new to machine learning and one of the things I don't understand about CNN is whether we have to learn the kernel values at every convolutional layer, or just learn a single set of kernel values ...
2
votes
1answer
24 views

Why convoloution neural net have to find filter values ?

I'm new to ML stuff and one of the thing that I don't understand about CNN, is that why CNN have to find the values of filter at convolution layer, why don' they use existing filters and only find the ...
0
votes
0answers
30 views

Training and validation loss increase with time in siamese network

I am trying to train a Siamese network using the approach described here but with a different architecture (similar to VGG-16). However, instead of using contrastive loss, I am using a Pdist loss. My ...
0
votes
0answers
19 views

Computation complexity and processing of one image for object detection in Convolutional Neural Network

How do I relate compute complexity in Convolutional Neural Network to processing time of one image in object detection for a given CPU/GPU's processing power? Say my CNN architecture needs ...
0
votes
3answers
52 views

Object localization with CNN

I am interested in locating the center of a playing card on the surface of a table: I have written a script so that I can generate images like this, where the card is moved around and rotated. My ...
1
vote
1answer
22 views

RNN vs Convolution 1D

Intuitively, are both RNN and 1D conv nets more or less the same? I mean the input shape for both are 3-D tensors, with the shape of RNN being ( batch, timesteps, features) and the shape of 1D conv ...
0
votes
1answer
27 views

Reduction of Feature map size in Convolutional Neural Network

In CNN, the way we reduce the feature map size at layers is we use pooling. Pooling makes feature map size into half. For the following network, if I want a new layer with feature map size somewhere ...
0
votes
0answers
18 views

Multiclass classification training data & validation

I am building a CNN based model for multiclass classification. There are close to 200 data points in my training data and there are a total of 30 classes for these 200 data points. I am a bit stuck ...
1
vote
0answers
35 views

Keras NN - loss gets stuck at 8.6791 [duplicate]

What does it mean when my neural network always gets stuck at the exact number 8.6791 when I use binary-crossentropy loss? Some strange local minimum? It happens regardless of my learning rate, ...
0
votes
0answers
8 views

Best practices of collecting & preparing data for video action/gesture recognition?

I am going to be collecting data to train a CNN/deep-learning model for gesture recognition. I have never worked in the domain of gesture recognition before, so I want to know if there are any tips or ...
4
votes
1answer
207 views

Back-propagation in Convolution layer

Most examples I found on the internet explain well back-propagation in convolution layer, but only with a single kernel and single input channel. I do not understand how to do back-propagation for ...
4
votes
0answers
31 views

Weight normalization technique used in Image Style Transfer

I am trying to implement the paper Image Style Transfer Using Convolutional Neural Networks. In section 2 - Deep image representations, the authors mention the following weight normalization technique:...
1
vote
0answers
12 views

I am trying to build a progressive auto encoder neural network and I am not sure how to discard old weights?

The goal of the network is simple, encode and decode images at a smaller scale and slowly increasing the network complexity, the input image size and its output quality. My current weights for my ...
2
votes
0answers
22 views

Lack of Batch Normalization Before Last Fully Connected Layer

In most neural networks that I've seen, especially CNNs, a commonality has been the lack of batch normalization just before the last fully connected layer. So usually there's a final pooling layer, ...
2
votes
0answers
26 views

comparing CNN vs other classification methods

I working on a classification problem. I have created Python code that takes certain labelled input data. This is then converted into two 2 dimensional arrays. The first array is an input array of ...
1
vote
0answers
19 views

deep learning, overfitting and identification problem

I have build a convolutional network with around "only" 1900 parameters for 4600 images in training set (observations) and I am still overfitting the training set. I view the problem like a system of ...
0
votes
1answer
24 views

Is there any sense to train deep conv net from the scratch after dataset changes?

I train deep conv model called resnet50 as the object detector. Time to time I make some changes in dataset or data augmentation. I usually use my last checkpoint to continue training on whole changed ...
2
votes
4answers
50 views

Difference between strided and non-strided convolution

conv = conv_2d (strides=) I want to know how non-strided convolution differs from strided i know how convolutions with strides work but don't know about the non-...
0
votes
1answer
43 views

What is the optimal number of neurons in fully connected layer in CNN?

I am reading this paper "Human Activity Recognition From Accelerometer Data Using Convolutional Neural Network". NN experts don't have to read the entire paper. I'm adding the important diagram here: ...