Questions tagged [convnet]

For questions regarding "Convolutional Neural Networks" (CNN)

152 questions with no upvoted or accepted answers
Filter by
Sorted by
Tagged with
4
votes
0answers
116 views

How to detect vanishing and exploding gradients with Tensorboard?

I have two "sub-questions" 1) How can I detect vanishing or exploding gradients with Tensorboard, given the fact that currently write_grads=True is deprecated in the Tensorboard callback as per "un-...
4
votes
0answers
53 views

Training value neural network AlphaGo style

I have been trying to replicate the results obtained by AlphaGo following their supervise learning protocol. The papers specify that they use a network that has two heads: a value head that predicts ...
4
votes
1answer
2k views

Gradient flow through concatenation operation

I need help in understanding the gradient flow through a concatenation operation. I'm implementing a network (mostly a CNN) which has a concatenation operation (in pytorch). The network is defined ...
4
votes
0answers
261 views

how to propagate error from convolutional layer to previous layer?

I've been trying to implement a simple convolutional neural network. But I've been stuck at this problem for over a week. To be specific, assume there are 3 layers in a convolutional pass, marked as ...
3
votes
0answers
71 views

What happens with activations?

I am playing with convolution network, assembling something between AlexNet and ResNet. Not very deep, about 10 conv. layers including 2 through residual connection, and 3 fully-connected layes at the ...
3
votes
0answers
32 views

Keras model with second to last sigmoid activated Conv1D layer followed by globalMaxPool outputs values outside [0,1]. Why?

I am trying to train a binary classifier. It is a residual network with skip layers etc. but ultimately, the bottom two layers are a 1D convolution with sigmoid activation followed by a global max ...
3
votes
1answer
142 views

Why is training and validation loss steadily rising (eventually to NaN) in this CNN of mine?

Dear ML and data scientists: I have 4 layers of gray scale images for every single biological specimen in my dataset. I am trying to train a 4-convolution CNN (see pytorch architecture below) to ...
3
votes
1answer
420 views

How to use PCA in CNN for image recognition using Keras?

I created a CNN model for image classification and I want to use Principal Component Analysis (PCA) but when I run pca.fit() code, the code still running for hours ...
3
votes
0answers
33 views

Computing derivatives for backpropagation across a convolution step

This will be a long post, but I hope it'll be instructive to anyone else in my position. I'm trying to find how the derivatives of the loss function are calculated with respect to the kernels and ...
3
votes
0answers
790 views

Multi label classification and sigmoid function

I'm new to neural networks so this may be silly question. I have build standard CNN network for image classification. I want multi-label classification network so I use binary_crossentropy as loss ...
3
votes
1answer
211 views

LeNet-5 - combining feature maps in C3 layer

Famous LeNet-5 architecture looks like this: The output of layer S2 has dimension: 10x10x6 - so basically an image with 6 convultions applied to it to derive features. If each dimension was again ...
3
votes
0answers
929 views

Cross validation for convolutional neural network

I am using Keras to create a CNN model, and I would to use K-fold cross-validation to train the dataset. The dataset contains images and I am using ...
3
votes
0answers
306 views

convolutional neural network with cross validation in Keras

I want to use K-fold cross-validation on my dataset of images. I am reading the data (images) from a directory. How do I use cross validation with convolutional neural network in Keras?
3
votes
1answer
41 views

Is there any work done on reconfigurable convolutional neural networks?

Convolutional Neural networks are used in supervised learning meaning models are always "set in stone" after training (architecture and paramters) so this might not even be possible, but is there any ...
3
votes
1answer
703 views

Keras bug NasNetlarge no top

I am trying to use NasNetlarge in Keras without the top but I cant get rid of the top: ...
3
votes
0answers
482 views

Keras - Masking CNNs

I have a 3D tensor on which I apply 2D convolutions. Sometimes, this 3D is padded both in width and height to have a fixed size. How could I apply masking (like with RNNs) so that the gradients ...
3
votes
0answers
630 views

Are non-zero paddings used?

I currently tried to figure out which paddings are directly supported by the frameworks: Tensorflow (tf.nn.conv2d): ...
2
votes
0answers
10 views

What is the meaning of 'concatenate' in this neural network architecture?

I am trying to understand the lane edge proposal network proposed in LaneNet for lane detection. My understanding of this is that a number of convolutional and pooling layers are first used to ...
2
votes
0answers
17 views

Precision Recall using Distance Matrix

Given a Pre-trained CNN model, I extract feature vectors for 3450 Reference (Winter) and 3450 Query images (Spring) and compare features with euclidean distance to plot the distance matrix besides ...
2
votes
0answers
21 views

Why does this implementation of SimpleNet use 3x3 kernels on it's final layer for cifar10?

Question; I'm trying to implement simplenet in tensorflow and I have a question that I can't seem to answer myself. The implementation I'm basing this off of is here: https://github.com/Coderx7/...
2
votes
0answers
38 views

Understanding the significance of LeNet-5 w/ MNIST data set

I'm beginning to learn about conv nets and started with what I understand to be one of the seminal works: LeNet-5. However, my limited experimentation doesn't seem to show any advantage over a single ...
2
votes
0answers
85 views

Why is convnet transfer learning taking so long?

I am using transfer learning to train a binary image classification model using keras' pretrained VGG16 model. The code can be found below : ...
2
votes
2answers
316 views

Model not learning when using transfer learning

I am working on a personal project on image classification (two classes) and am trying to see how the MobileNet v2 structure would perform. While training the training accuracy is already quite high ...
2
votes
0answers
35 views

Did I do the right thing in my CNN Keras (class imbalance - augmentation)

To implement my Binary CNN in keras, I had a dataset of ~~35000 images but only 700 is from one class and all the others are from the other class, so what I did: I get the 700 unique images from class ...
2
votes
0answers
436 views

1x1 convolutions, equivalence with fully connected layer

I'm confused by the concept of equating a 1x1 convolution with a fully connected layer. Take the following simple example of a 1x1 convolution of 2 input channels each of size 2x2, and a single output ...
2
votes
0answers
40 views

CNN kernel location for input image

Given a CNN, say AlexNet: How could one relate kernel locations at the 3rd conv block, i.e 13x13 filter size to the input image. Would that give a meaningful representation in terms of the input ...
2
votes
0answers
267 views

Data augmentation / feature extraction on pre-trained convnets

I'm reading 'Deep Learning with Python' by François Chollet, which is an excellent book. He talks about using pre-trained convnets (in his example, VGG16) and then running smaller datasets to tweak ...
2
votes
0answers
379 views

Tensorflow CNN sometimes converges, sometimes not

originally asked on stackoverflow, deleted Im having trouble traing a convolutional neural network in tensorflow. When I start my program, sometimes the model learns nicely (cost/cross_entropy goes ...
2
votes
1answer
258 views

MLP conv layers

When should MLP conv layers be used instead of normal conv layers? Is there a consensus? Or is it the norm to try both and see which one performs better? I would love to better understand the ...
2
votes
0answers
23 views

Contextual Object Detection

An example of what I'd like to do is identify the price of a product on a product page. While I can train a CNN to identify prices, it's likely that it would recognise every instance of a price on a ...
2
votes
0answers
795 views

How to detect multiple digits in one image using CNN

The network in CNN recognize one digit for one image, how can I recognize two or more digits in one image. A proper way I've thought about is modifying the weights in the FC layer. In other words, ...
2
votes
0answers
125 views

What does it mean if performance of two different iterations of the same network (CNN model) varies a lot?

So I trained CNN model for people detection on caltech-pedestrian dataset: Then I was curious and evaluated the model in every 1000th iteration on Evaluation toolbox(I guarantee, there is no bug in ...
2
votes
0answers
91 views

ConvNet for Archery App

I'm looking at trying to build a model to score an archery end from a photo but i'm not sure about the optimal ConvNet architecture. What output neurons would I use given I can have an unknown number ...
2
votes
0answers
102 views

What are natural (computed) pre-images useful for?

I've just been reading Zeiler, M.D. and Fergus, R., 2014, September. Visualizing and understanding convolutional networks. In European Conference on Computer Vision (pp. 818-833). Springer ...
2
votes
0answers
78 views

Discrepancy in probability calculations in paper 'Multi-digit Number Recognition…'

In the paper, 'Goodfellow, I., et al. Multi-digit Number Recognition from Street View Imagery using Deep Convolutional Neural Networks. ICLR, 2014', on page 10 there is a table which calculate $\log(P(...
2
votes
2answers
291 views

How to determine the number of the training images in Keras after data augmentaion?

I want to create a CNN model and I am using data augmentation. I want know the number of augmented images in Keras. How to determine the number of the training images in Keras after data augmentation?...
1
vote
0answers
31 views

Why does regression model predict the same output

I have built a CNN for regression, but it is giving identical predictions (up to 8 sig. fig.) for almost 1/3 of the test data set. (The other outputs are different.) Is there a reason why this might ...
1
vote
0answers
26 views

How to interpret Deep learning network architecture into a diagram?

How to draw this Deep learning network architecture diagrams? I'm using Faster R-CNN: R50-FPN. Any ideas or tip to convert this to a diagram? Or just to know which are input, hidden and output ...
1
vote
0answers
47 views

Gender Prediction from Offline Handwriting Using Convolutional Neural Networks

Starting from the fact that handwritten documents style are gender-dependent (male and female have different writing styles), I'm trying to predict writer's gender from its handwritten scripts using ...
1
vote
0answers
13 views

Using a 3-D convolutional layer to simulate a 2-D convolutional layer

I've asked this question on the AI StackExchange , but I received no insight so I'd like to ask it here. Is using a filter of size (1, x, y) on a 3-D convolutional layer the functionally equivalent ...
1
vote
0answers
13 views

The idea behind Generalized Max Pooling

I am trying to understand the idea of "Generalized Max Pooling". It seems they try to make the 'pooled' representation similar to the features. If so I feel some rare discriminating features could ...
1
vote
0answers
19 views

Preparing text for modeling in dialogue structure

I'm working on implementing the DialogueGCN code from this paper. Its a model that classifies the 'emotion' from utterances of text within a conversation. As this model takes into account speaker ...
1
vote
0answers
16 views

Train a competitive layer on nonnormalized vectors using LVQ technique

How can we train a competitive layer on non-normalized vectors using LVQ technique ? The net input expression for LVQ networks calculates the distance between the input and each weight vector ...
1
vote
0answers
36 views

Why when I apply GaussianBlur in my images, my model overfit? CNN KERAS

I have 1400 images (700 each class), and i'm using vgg-16 to classificate between one and other class. But when I apply the preprocess method Gaussian Blur (which seem to be very much clear to see the ...
1
vote
0answers
17 views

What is the difference between these 2 training scenarios?

I have a very large dataset and due to computational constraints, I have to divide the data into 20 parts (each part is around 1.5GB). I constructed a deep CNN model using Keras for this dataset. The ...
1
vote
0answers
60 views

Shuffling the images with DataImageGenerator and flow_from_directory in a Medical application (tensorflow 2/Keras)

I am facing the following situation: I have CT images (scans) of patients where 10 images describe each patient. These images are stored in separate directories based on what is the focus of each ...
1
vote
0answers
17 views

What is the use of having shared weights in later layers of a CNN?

In a CNN, all the neurons in a single layer use the same weights and bias. As a result, all the neurons detect the same feature. The early layers of a CNN detect simple features like edges and hence ...
1
vote
0answers
31 views

When do I use Multiply and Add

I want to know the effect of Add and Multiply in keras by functionality. The dumb way of thinking is that they are meant to add ...
1
vote
0answers
37 views

How to backpropogate error from convolutional layer with respect to the input when using multiple channels

I have been attempting to implement a Convolutional Neural Network in python and have run into a bit of a roadblock. When backpropogating the error in a convolutional layer let us say that we receive ...
1
vote
2answers
38 views

ConvNet with concatenated data

I have a basic question regarding convolutional neural network. Assume I have a set of 1000 RGB images and I train a CNN from this set. I can obviously split each of my RGB images into 3 different ...