Questions tagged [convolutional-neural-networks]
For questions about convolutional neural networks, also known as CNN or ConvNet.
741
questions
0
votes
0answers
5 views
Indoor scene understanding dataset which can be used commercially?
Is there a (large) indoor scene understanding datasets (providing instance segmentation masks) which can be used commercially ?
All large scene understanding datasets (SceneNet, ScanNet, InteriorNet, ....
2
votes
1answer
28 views
Is it possible to apply the associative property of the convolution operation when it is followed by a non-linearity?
The associative property of multidimensional discrete convolution says that:
$$Y=(x \circledast h_1) \circledast h_2=x\circledast(h_1\circledast h_2)$$
where $h_1$ and $h_2$ are the filters and $x$ is ...
1
vote
0answers
16 views
How to measure/estimate the energy consumption of CNN models during testing?
Does someone know a method to estimate / measure the total energy consumption during the test phase of the well-known CNN models? So with a tool or a power meter...
MIT has already a tool to estimate ...
1
vote
1answer
25 views
Can fully connected layers be used for feature detection?
I need help in understanding something basic.
In this video, Andrew Ng says, essentially, that convolutional layers are better than fully connected (FC) layers because they use fewer parameters.
But I'...
1
vote
0answers
39 views
When to use convolutional layers as opposed to fully connected layers?
I am still new to CNNs, but I would like to check my understanding between when to use convolutional layers versus fully connected layers.
From what I have read, we can use convolutional layers with ...
1
vote
0answers
28 views
Atari Games: Pretrained CNN to accelerate training?
DQN for Atari takes considerable training time. For example, the 2015 paper in Nature notes that algorithms are trained for 50 million frames or equivalently around 38 days of game experience in total....
0
votes
1answer
14 views
How is it possible to get the output size of `n` Consecutive Convolutional layers? [closed]
Given network architecture, what are the possible ways to define fully connected layer fc1 to have a generalized structure such as ...
0
votes
0answers
21 views
DQN not learning the game Pong
Hello I'm trying to implement DQN Agent to play Atari-Pong game. But still after 500.000 parameter updates, the model still scores -21. This is my code, do you have any idea what could be wrong?
...
0
votes
0answers
10 views
CNN keras accuracy not improving
I am trying to duplicate and learn from example given on this website . With my little modification, I am trying to simple exchange color for example like red to orange in an image. The original ...
0
votes
0answers
9 views
Is there a 2D (sampled frequencies) music network in some kind of zoos like Imagenet to try it in order to get style transfer?
It looks like the music which is sampled and each sample has been transferred into frequency domain with FFT is in fact 2D object, capable to be dealed with CNN networks. So, it looks like if some ...
0
votes
0answers
16 views
Binary classification to recognize blobs on pictures generates many false-positive results
I am training a NN for blobs vs non-blobs recognition.
Blobs example:
Non-blobs:
Keras architecture is:
...
1
vote
1answer
20 views
How will the input be preserved as we go deeper in CNN, where dimensions decrease drastically?
Our length of our like feature representation actually decreases as we go deeper into the CNN, I mean to say that horizontal and vertical lengths decrease while depth(channels) increase. So, how will ...
0
votes
0answers
6 views
NN for defect detection
New to NN, I'm looking to get advice for the architectural implementation using tensorflow of a neural net for defect detection in the material as well as suggested image preprocessing to improve NN ...
0
votes
0answers
13 views
How much data do we need for making a successful de-noising auto-encoder?
Is there a guide how much data do you need for making successful denoising model using autoencoders?
Or the rule is, the more data, the better it is?
I tried with small dataset 350 samples, to see ...
1
vote
0answers
10 views
How is the performance of a CNN trained with monochrome images on image recognition tasks degraded?
For CNN image recognition tasks, like object recognition/face recognition/object segmentation/posture recognition, are there experiment results about how much will the performance be degraded with ...
1
vote
2answers
21 views
How can I have the same input and output shape in an auto-encoder?
I'm building a denoising autoencoder. I want to have the same input and output shape image.
This is my architecture:
...
0
votes
1answer
23 views
What is meant by the number of channels of a network?
Currently, I am reading Rethinking Model Scaling for Convolutional Neural Networks. The authors are talking about a different way of scaling convolutional neural networks by scaling all dimensions ...
0
votes
0answers
10 views
Optimal RL agent's representation of 3D-grid data: 2D Slices and CNN encoding. Suggestions?
environment
My agent needs to navigate in a 64x64x64 discrete 3d grid environment, and remove certain voxels. Voxels can be in a number of states: should-remove, should-not-remove, empty. It can move ...
2
votes
1answer
32 views
Does the number of parameters in a convolutional neuronal network increase if the input dimension increases?
If I have a convolutional neuronal network, does the input dimension change the number of parameters? And if yes, why? If the sizes and lengths of the filters are still the same, how can the number of ...
0
votes
0answers
9 views
Are there references for real-time 3D shape measurement using convolutional neural network?
I want to develop a real-time 3D shape measurement using convolutional neural network (CNN) for vehicle. Could you recommend me references for that? Thanks.
0
votes
0answers
21 views
Normalization of every patch of CNN input separately
What would be the effect of normalizing each input patch going to Convolutional layer separately.
Let's say our input is 64 channels of the size 224x224 (like is the case for some hidden layers in ...
1
vote
2answers
85 views
Do convolutional neural networks perform convolution or cross-correlation?
Typically, people say that convolutional neural networks (CNN) perform the convolution operation, hence their name. However, some people have also said that a CNN actually performs the cross-...
0
votes
0answers
25 views
How to train neural network a math multiplication table? [migrated]
I am trying to train neural network (brain.js) a multiplication table. It is not going too well: requires lots of hidden layers, iterations and very small error threshold, and the results are still ...
0
votes
1answer
60 views
What is a convolutional neural network?
Given that this question has not yet been asked on this site, although similar questions have already been asked in the past (e.g. here or here), what is essentially a convolutional neural network (...
1
vote
1answer
42 views
How do GPUs faciliate the training of a Deep Learning Architecture?
I would love to know in detail, how exactly GPUs help, in technical terms, in training the deep learning models.
To my understanding, GPUs help in performing independent tasks simultaneously to ...
1
vote
0answers
20 views
Overcome caveats on using Deep Learning for faster inference on limited performance availability
I am working in the field of Machine Vision, where accuracy and performance both play a major factor in deciding the approach towards a problem. Traditional rule based approaches work quite well in ...
1
vote
0answers
29 views
What is the reason for different learned features in upper and lower half in AlexNet?
I was reading AlexNet paper and the authors quoted
the kernels on one GPU were "largely color agnostic," whereas the kernels on the other GPU were largely "color-specific."
The upper GPU takes ...
1
vote
1answer
64 views
What is a fully convolution network?
I was surveying some literature related to Fully Convolutional Networks and came across the following phrase,
A fully convolutional network is achieved by replacing the parameter-rich fully ...
1
vote
2answers
51 views
Can we achieve what a CNN can do with just a normal neural network?
When I was learning about neural networks, I saw that a complex neural network can understand the MNIST dataset and a simple convolution network can also understand the same. So I would like to know ...
1
vote
0answers
19 views
How to calibrate model's prediction given past images?
I want to predict how open is the mouth given a face image. It's a regression problem (0= mouth not open, 1=mouth completely open). And something between 0 and 1 is also allowed. ConvNet works fine ...
2
votes
2answers
82 views
How to convert sequences of images into state in DQN?
I recently read the DQN paper titled: Playing Atari with Deep Reinforcement Learning. My basic and rough understanding of the paper is as follows:
You have two neural networks; one stays frozen for a ...
1
vote
0answers
92 views
Can we force the initial state of a neural network to produce an “unknown” class?
Has anyone investigated ways to initialize a network so that everything is considered "unknown" at the start?
When you consider the ways humans learn, if something doesn't fit a class well enough, it ...
0
votes
0answers
18 views
How can I build a model that replaces a feature of one image with another feature?
I would like to build a neural network (using TensorFlow) that is able to take two animals, and replace a feature in the second with one in the first. For example, if given a dog and cat, the cat's ...
0
votes
0answers
7 views
Identifying and Labeling multiple letters in image
While I attempt to learn AI/ML I have taken on the task to create a Boggle solver. The idea is that a system could take an image of a Boggle arrangement of letters and identify the letters (and the ...
3
votes
2answers
155 views
What is the effect of using pooling layers in CNNs?
I know how pooling works, and what effect it has on the input dimensions - but I'm not sure why it's done in the first place. It'd be great if someone could provide some intuition behind it - while ...
1
vote
0answers
19 views
Deep learning techniques with time-fixed, time-dependent and imaging data
I have a question about the use of deep learning techniques with time-fixed features and images (setting 1) and time-dependent features (setting 2). (I am pretty new to the deep learning world so ...
1
vote
0answers
28 views
Which one is more important in case of different loss optimization algorithms, Speed or the Route?
We have different kinds of algorithms to optimize the loss like AdaGrad, SGD + Momentum, etc. Some are more commonly used than the others. In some algorithms, they usually range out before they ...
0
votes
0answers
20 views
Regularization to enforce feature learning
Is there any research into ways to enforce feature selection across classes by network structure?
Given the number of parameters in NN, even convnets are prone to over fitting.
I'm curious if there ...
1
vote
0answers
20 views
Banding artifacts in CNN
I was working on a CNN for HDR image generation from LDR images. I used an encoder-decoder architecture and merged the input with the decoder output. However I'm getting some banding artifacts in the ...
2
votes
1answer
107 views
Why do we need convolutional neural networks instead of feed-forward neural networks?
Why do we need convolutional neural networks instead of feed-forward neural networks?
What is the significance of a CNN? Even a feed-forward neural network will able to solve the image classification ...
1
vote
0answers
18 views
How can I find the similar non-zero connections between different levels of sparsity of the same network?
I am pruning a neural network (CNN and Dense) and for different sparsity levels, I have different sub-networks. Say for sparsity levels of 20%, 40%, 60% and 80%, I have 4 different sub-networks.
Now, ...
1
vote
0answers
11 views
Is there a difference between using 1d conv layers and 2d conv layers with kernel with size of 1 along other than time dimension?
Let's assume I use convolutional networks for time-series prediction. Data I feed to the network have 1 channel depth, height of number of periods and number of features is the width, so the frame ...
0
votes
1answer
23 views
What is the advantage of using Google's Coral over Nvidia's Xavier?
I was reading about the possibility of using Google's Coral for deep learning-based object detection and image classification. I heard it has a good speed in terms of frames/sec.
I also read that ...
1
vote
0answers
21 views
How do we choose the filters for the convolutional layer of a convolution neural network?
Since the hidden layers of a CNN work as a trainable feature extractor, more detailed content based on a larger number of pixels shall require bigger filter sizes. But for cases where localized ...
2
votes
0answers
17 views
How does sampling works in case of imbalanced image datasets?
I am solving a problem of image classification of the image dataset for 3 classes. Dataset is highly imbalanced.
How will sampling (either over- or under-sampling) work in that case? Should I remove (...
1
vote
0answers
9 views
How do CNNs or RNNs “stack the feature of nodes by a specific order”?
I am trying to understand the following statement taken from the paper Graph Neural Networks: A Review of Methods and Applications (2019).
Standard neural networks like CNNs and RNNs cannot handle ...
2
votes
0answers
20 views
How is visual attention mechanism different from a two branch convolutional neural network?
I am doing some research on the visual attention mechanism in remote sensing domain (where the features learnt from one layer are highlighted using the attention mask derived from another layer). From ...
3
votes
1answer
96 views
How do I calculate the partial derivative with respect to $x$?
I am trying to implement CNN using python Numpy.
I searched so much, but all I found was for one filter with one channel for Convolution.
Suppose we have an X as Image with this shape: ...
1
vote
1answer
50 views
If features are always positives, why do we use RELU activation functions?
When does it happen that a layer (either first or hidden) outputs negative values in order to justify the use of RELU?
As far as I know, features are never negative or converted to negative in any ...
3
votes
0answers
16 views
Conversion of strided filter gradient to convolutional form
I'm implementing strided 2D convolution. My formula looks like this:
$$y_{i, j} = \sum_{m=0}^{F_h - 1}\sum_{n=0}^{F_w - 1} x_{s\cdot i + m, s\cdot j + n}\,f_{m, n}, \tag{1}$$ where $s$ is the stride
(...
