Questions tagged [convolutional-neural-networks]
For questions about convolutional neural networks, also known as CNN or ConvNet.
804
questions
2
votes
1answer
39 views
Can we apply transfer learning between any two different CNN architectures?
There are many types of CNN architectures: LeNet, AlexNet, VGG, GoogLeNet, ResNet, etc. Can we apply transfer learning between any two different CNN architectures? For instance, can we apply transfer ...
1
vote
0answers
19 views
Pytorch Deep q network not learning and step not stepping towards target
I am trying to create a simple deep q network for rl with conv2d layers. I can’t figure out what I am doing wrong, and the only thing I can see that doesn’t seem right is when I get the model ...
2
votes
1answer
32 views
Should I remove the text overlaying some images in the dataset before training the CNN?
If I am attempting to train a CNN on some image data to perform image classification, but some of the images have pieces of text overlaying them (for the purpose of description to humans), then is it ...
0
votes
0answers
20 views
How to calculate GPU memory requirements for ultra-high-resolution image classification? [closed]
I am working on deep CNN image classification for images which are natively about 3200x3200 at full resolution. Although our research will explore downsampling and windowing methods, we also want to ...
1
vote
0answers
18 views
Duplicating calculations in CNN-LSTM architecture
I want to use frames from video game and analyze them using CNN and LSTM. But when I have the model defined like that
...
1
vote
1answer
31 views
Research paths/areas for improving the performance of CNNs when faced with limited data
I've been reading through the research literature for image processing, computer vision, and convolutional neural networks. For image classification and object recognition, I know that convolutional ...
0
votes
1answer
21 views
How to construct input dependent convolutional filter?
I am constructing a convolutional variational autoencoder for images, starting out with mnist digits. Typically I would specify convolutional layers in the following way:
...
0
votes
2answers
35 views
How could I convolve a 4D image and a 4D filter with stride?
I want to create a CNN in Python, specifically, only with NumPy, if possible. For optimizing the time of convolution (actually correlation) in the network, I wanna try to use FFT-based convolution. ...
-2
votes
0answers
63 views
How do I write the architecture (layers, activation functions, etc.) of a neural network in pseudocode?
I am looking to convert a CNN model written in Python (keras) to pseudocode. I am mostly trying to find out the logic on how to describe the layers, the filters, the activation functions etc in ...
1
vote
1answer
34 views
Are mult-adds and FLOPs equivalent?
I am comparing different CNN architectures for edge implementation. Some papers describing architectures refer to mult-adds, like the MobileNet V1 paper, where it is claimed that this net has 569M ...
0
votes
1answer
29 views
How deep convolution filters should be?
In ConvNet, after having decided the number of layers for it, one has to decide the filter size and depth for each layer. The intuition behind the spatial filter size is the number of pixels in the ...
0
votes
1answer
29 views
Should image augmentation be applied before or after image resizing?
For the purposes of training a Convolutional Neural Network Classifier, should image augmentation be done before or after resizing the training images?
To reduce file size and speed up training time, ...
0
votes
0answers
13 views
fastai- how to visualize each conv2d layer in resnet50 [migrated]
I am using fastai v1 and trained a resnet50 model on my image data set.
Though I want to visualize how the activations look like at each layer or at least some of the intermediate layers. I found one ...
2
votes
2answers
134 views
How to identify if 2 faces contain the same person?
I have got numerous frames and I've detected all the faces in all the frames using Retinaface. However I need to track the faces of people over frames.
For this purpose, I assumed I could try finding ...
0
votes
1answer
42 views
Are there neural networks with 3-dimensional topologies?
The topologies (or architectures) of the neural networks that I have seen so far are only 2-dimensional. So, are there neural networks whose topology is 3-dimensional (i.e. they have a width, height, ...
1
vote
1answer
26 views
Why does CNN forward pass take longer compared to MLP forward pass? [closed]
Let's take a 32 x 32 x 3 NumPy array and convolve with 10 filters of size 2 x 2 x 3 with stride 2 to produce feature maps of volume 16 x 16 x 10. The total number of operations - 16 * 16 * 10 * 2 * 2 *...
1
vote
1answer
46 views
What's a good neural network for this problem?
I am very new to the field of AI so please bear with me.
Say there is a dice with three sides, -1,0 and 1, and I want to predict which side it lands on (so only one output is needed I guess). The ...
0
votes
1answer
26 views
What is the dimension of my output of the form (2n + 1, 2n + 1, #filters) after a MaxPooling layer
I'm trying to white board the different mechanisms behind a convolutional neural network.
I have on question regarding the dimension of my volume after using a max pooling layer.
Let's suppose I have ...
2
votes
0answers
19 views
How is the data labelled in order to train a region proposal network?
I don't get how the training of the RPN works. From the forward propagation, I have $W \times H \times k$ outputs from the RPN.
How is the training data labeled such that I can use the loss function ...
0
votes
2answers
24 views
Stride in convolutional neural networks (horizontal/vertical direction)?
I feel like this question has been asked before, but I can't seem to find any old material on this.
In the convolutional layer for CNNs, when you specify the stride of a filter, typical notes show ...
0
votes
0answers
26 views
Keeping track of multiple faces throughout a video
I have a video where multiple persons are seated. I need to keep track of the emotions they show throughout the video. My final result should be a csv file with all the emotions depicted by each ...
0
votes
0answers
17 views
Class activation maps for 3D Convolutional neural network?
I have implemented a 3D convolutional neural network and I was not able to find resources for interpretation for my model. I have found some techniques such as GradCam and GradCam++ but these generate ...
0
votes
0answers
10 views
How to combine specific CNN models that work better at slightly different tasks?
I'm not sure how to describe this in the most accurate way but I'll give it a shot.
I've developed a Inception-Resnet V2 model for detecting audio signals via spectrogram. It does a pretty good job ...
0
votes
0answers
29 views
Can I use the full dataset for training the CNN feature extractor?
I built a CNN model for extracting features from its dense layer. The extracted features are then used for classification using KNN and Random Forest classifier.
My question is, can I use the whole ...
1
vote
1answer
26 views
Why is the convolution layer called Conv2D?
When I build a convolution layer for image processing, the filter parameters should have 3 dimensions, (filter_length, filter_width, color_depth) is that correct?
...
0
votes
0answers
18 views
Object detection using CNN model architectures
I've used LabelImg to create labels for my images using YOLO. After that, I would like to input the images and labels into a CNN model, like a VGG or ResNet. I've searched a lot and have not found ...
1
vote
0answers
13 views
What's the nearest neighbor algorithm used for upsampling?
Additionally, by default, the UpSampling2D layer will use a nearest neighbor algorithm to fill in the new rows and columns. This has the effect of simply doubling rows and columns, as described and is ...
0
votes
0answers
13 views
Why it is reshaped the last layers of VGG_UNet segmentation model?
I want to do a multiclass segmentation task using deep learning (in python). Here, is a summary of vgg_unet model that is mainly collected from GitHub. So, in my dataset 8 labels are available. So, at ...
3
votes
0answers
27 views
What does “convolve k filters” mean in the AlphaGo paper?
On page 27 of the DeepMind AlphaGo paper appears the following sentence:
The first hidden layer zero pads the input into a $23 \times 23$ image, then convolves $k$ filters of kernel size $5 \times 5$ ...
0
votes
1answer
32 views
How can I implement 2D CNN filter with channelwise-bound kernel weights?
I would like to bind kernel parameters through channels/feature-maps for each filter. In a conv2d operation, each filter consists of HxWxC parameters I would like to have filters that have HxW ...
0
votes
0answers
17 views
Conditional GAN on harder datasets
I have seen conditional GANs often applied to easier datasets like MNIST and CIFAR-10 to reasonable success, but at the same time these datasets are simple enough that naïve CNNs can fairly easily max ...
0
votes
0answers
20 views
Faster RCNN-RPN Network Training
I am trying to understand RPN network in Faster RCNN.
I understand the concept of RPN network,
Pass the input images to the pre trained CNN, and get the output as feature maps
Make fixed size of the ...
1
vote
1answer
37 views
Do all filters of the same convolutional layer need to have the same dimensions and stride?
In Convolutional Neural Networks, do all filters of the same convolutional layer need to have the same dimensions and stride?
If they don't, then it would seem the channel produced by each filter ...
1
vote
1answer
46 views
What is the time complexity of the upsampling stage of the U-net?
I am trying to determine the complexity of the neural network we use. The neural network is a U-net generator with an input shape of NxN (not an image but image-like data) and output of the same shape....
1
vote
1answer
151 views
What is the computational complexity of the forward pass of a convolutional neural network?
How do I determine the computational complexity (big-O notation) of the forward pass of a convolutional neural network?
Let's assume for simplicity that we use zero-padding such that the input size ...
0
votes
0answers
14 views
Limbs for PAFs in OpenPose
What limbs are used in openpose for the PAF? If you look at the skeletal reconstruction one would assume it is $ears - eyes$, $eyes - nose$, $nose-neck$, $neck - hips$, $shoulders - elbows$, $elbows- ...
0
votes
0answers
17 views
Image classification on SVG format
To best of my knowledge, images are usually fed in pixel format to ML models.
Is there any work that does image classification where the image format is SVG?
0
votes
0answers
14 views
Does the selective search algorithm in object detection learn?
I am trying to get a better grasp of how object detection works. I (almost) completely understand the concept behind RPNs. However I am little bit confused with the selective search algorithm part. ...
0
votes
0answers
15 views
Is RCNN resolution-independent, if keeping feature size constant?
From what I understand, (Faster/Mask-)RCNN is fully convolutional. The backbone is fully convolutional, and the region proposal network (RPN) creates anchors on the feature map with a fixed stride.
...
0
votes
0answers
24 views
Is Bayesian NN vs adding random data more accurate?
I’m trying to train a classifier to recognize if people are wearing seatbelts. What if the person submitted a picture unrelated to a seatbelt classifier?
Would I create an image label that is full of ...
0
votes
0answers
18 views
The convolutional network architectures with enhanced invariance
It is well known, that CNN have advantage with respect to the Dense neural networks in the image classification and other pattern recognition tasks, because they have a translationall invariance built ...
0
votes
0answers
28 views
Since both RoI Align and PrRoI Pooling use bilinear interpolation, why is RoI Align discrete while PrRoI Pooling continuous?
I have two questions.
Since both use bilinear interpolation, why is RoI Align discrete while PrRoI Pooling continuous?
Could anyone explain the intuition behind the derivative of PrPool()?
0
votes
2answers
47 views
How to handle images that don’t pertain to image classifier at all?
I am trying to create a CNN model that classifies if a person is wearing a seatbelt or not to verify they drive safely. I know to get images of people wearing seatbelts and people not wearing ...
0
votes
0answers
34 views
How to choice CNN architecture for stitching images
I decided to start learning neural networks by creating a bot for the game. One of the intermediate steps is to create a global map from a series of inaccurate overlapping sub-maps. This task can be ...
0
votes
0answers
47 views
ConvNet - What to improve regarding architecture, procedure and technique?
I have a dataset of 180k images of license plates (so, not necessary to localize the license plate at first) for which I try to recognize the characters on the images (License plate recognition). All ...
0
votes
0answers
51 views
What are some scalable approaches to perform anomaly detection (for images with small cracks) with unsupervised learning?
I have some images with anomalies, like small cracks, but it's an imbalanced dataset.
Please, suggest some effective scalable approaches. Should I consider convolutional auto-encoders? It's supposed ...
1
vote
0answers
47 views
Extending patch based image classification into image classification
I am trying to classify tampered, pristine images from set of images, in that I have built a network in which I would divide the image into multiple overlapping patches and then classify them into ...
0
votes
0answers
44 views
Non-trainable regularizer in loss function
I train a fully convoluted network for semantic segmentation. To each convolution blocks, I associate a module pruning feature maps to reduce the quantity of information generated by the network. From ...
3
votes
1answer
58 views
Can residual neural networks use other activation functions different from ReLU?
In many diagrams, as seen below, residual neural networks are only depicted with ReLU activation functions, but can residual NNs also use other activation functions, such as the sigmoid, hyperbolic ...
4
votes
2answers
103 views
What should I do with the flatten layer during back-propagation?
I'm creating a CNN network without other frameworks such as PyTorch, Keras, Tensorflow, and so on.
During the forward pass, the Flatten layer reshapes the previous ...
