Questions tagged [convnet]

For questions regarding "Convolutional Neural Networks" (CNN)

Filter by
Sorted by
Tagged with
4
votes
3answers
5k views

Number of Fully connected layers in standard CNNs

I have a question targeting some basics of CNN. I came across various CNN networks like AlexNet, GoogLeNet and LeNet. I read at a lot of places that AlexNet has 3 Fully Connected layers with 4096, ...
4
votes
1answer
2k views

Why is the learning rate for the bias usually twice as large as the the LR for the weights?

I've noticed in a few caffe models I've been working with that the learning rate for the bias is often set to be twice that of the one for the weights. Another user mentions that this is the case in ...
4
votes
3answers
4k views

Should there be a flat layer in between the conv layers and dense layer in YOLO?

Should there be a flat layer in between the conv layers and dense layer in YOLO? It's something not specified in the paper, but I see most implementations of YOLO on github do this. In my ...
5
votes
1answer
772 views

Convolution Neural Network Loss and performance

I have a set of about ~100,000 training examples. Ratio of positive to negative example is roughly 1:2. The true ratio is more like 1:100 so this represents a major downsampling of the negative class. ...
2
votes
1answer
1k views

What is the best architecture for Auto-Encoder for image reconstruction?

I am trying to use Convultional Auto-Encoder for its latent space (embedding layer), specifically, I want to use the embedding for K-nearest neighbor search in the latent space (similar idea to ...
1
vote
2answers
41 views

Is Flatten() layer in keras necessary?

In CNN transfer learning, after applying convolution and pooling,is Flatten() layer necessary? I have seen an example where after removing top layer of a vgg16 ,first applied layer was ...
4
votes
2answers
164 views

Why are results without Transfer Learning better than with Transfer Learning?

I developed a neural network for license plate recognition and used the EfficientNet architecture (https://keras.io/api/applications/efficientnet/#efficientnetb0-function) with and without pretrained ...
5
votes
5answers
2k views

How to fix these vanishing gradients?

I am trying to train a deep network for twitter sentiment classification. It consists of an embedding layer (word2vec), an RNN (GRU) layer, followed by 2 conv layers, followed by 2 dense layers. Using ...
0
votes
1answer
15 views

Understanding declared parameters in my Conv2d layer of my convolutional neural network

I am trying to understand the architecture of my keras model implemented by the sequential model. Here is a piece of the code : ...
0
votes
0answers
11 views

Is it possible to create a 3d to 2d U-net?

I was curious if it is possible to create a U-net type architecture that takes in a 3d image and outputs a 2d image? Or, alternatively, would some other architecture be better suited for this problem? ...
12
votes
2answers
7k views

Document classification using convolutional neural network

I'm trying to use CNN (convolutional neural network) to classify documents. CNN for short text/sentences has been studied in many papers. However, it seems that no papers have used CNN for long text ...
20
votes
5answers
8k views

Convolutional neural network overfitting. Dropout not helping

I am playing a little with convnets. Specifically, I am using the kaggle cats-vs-dogs dataset which consists on 25000 images labeled as either cat or dog (12500 each). I've managed to achieve around ...
222
votes
11answers
208k views

What are deconvolutional layers?

I recently read Fully Convolutional Networks for Semantic Segmentation by Jonathan Long, Evan Shelhamer, Trevor Darrell. I don't understand what "deconvolutional layers" do / how they work. The ...
4
votes
0answers
4k views

Memory problems with smaller CNN [closed]

Hello everyone I'm having a weird problem. I having problem with one of two models I've been using. Models take an image data as input and outputs joystick and keyboard information. A simple CNN, ...
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 ...
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 ...
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 ...
0
votes
0answers
37 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 ...
1
vote
2answers
937 views

How to merge two CNN deep learning model using weighted sum and weighted product in Keras?

I am using Keras to create a deep learning model and I want to merge two CNNs by using weighted sum or weighted product. How can I merge two CNNs using weighted sum and weighted product?
0
votes
0answers
20 views

Plotting scikit-learn confusion matrix returns no values in the last class

I am attempting to create a confusion matrix using Scikit-Learn for a multiclass classification CNN, and it works well except for the fact that it does not provide ...
1
vote
1answer
17 views

pytorch convolution with 0-stride along one dimension

For some square images, I'd like to use torch.nn.Conv2d with the kernel as a vertical block. As in, the kernel size is defined as max value of the first dimension by 1. Since the first dimension has ...
0
votes
1answer
21 views

Varying Image sizes in Tensorflow Malaria dataset | Dealing with unclean tensorflow data

I am trying to build a CNN based image recognition system for the Tensorflow malaria dataset. I loaded the dataset (~27k RGB images) using conventional tensorflow_datasets syntax. After some data ...
0
votes
1answer
26 views

Is always converting a input vector into matrix and apply cnn's good idea?

I know the benefits of using cnns(reduced size weight matrices). Is it a good idea to convert a input vector(which is not a image) into a matrix and apply cnn's. What I understand is that it should ...
0
votes
0answers
19 views

How to use AutoEncoders or VAE for initializing custom CNN architecture?

I would like to use AutoEncoder or VAE in order to learn set of features which I can use to initialize training procedure of a custom CNN architecture that I've build. Here is the code: ...
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 ...
2
votes
1answer
38 views

Back propagation through a simple convolutional neural network

Hi I am working on a simple convolution neural network (image attached below). The input image is 5x5, the kernel is 2x2 and it undergoes a ReLU activation function. After ReLU it gets max pooled by a ...
0
votes
0answers
33 views

1D convolutional neural network validation improvement

I created 1D CNN in Keras, but I'm having issues with validation loss and accuracy. I have 24k records, 22 features. Is my model overfitting or what is going on so validation loss and accuracy is ...
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 ...
0
votes
2answers
361 views

Keras Conv1D model Input_shape value error

I am not sure why I am receiving this value error. Additionally, I haven't found a tutorial that explicitly talks about the appropriateness of size of filters and kernel. I would appreciate some input ...
0
votes
0answers
6 views

How to use Meta-SGD with U-Net

I'm guessing how to use Meta-SGD with U-Net network. I've been searching for an example, but I haven't found anything yet. I'm reading the book Hands On Meta Learning with Python, which has the ...
1
vote
1answer
58 views

Reason for Training and test loss sudden increment after some epochs keras

We know that if training and test loss are different from each other, our model is over-fitting. However, if both get high after some epochs, how can we justify it? One way to solve it is to reduce ...
6
votes
3answers
13k views

How to improve loss and avoid overfitting

I'm trying to build a 2 class image classifier using the architecture suggested in first part of this blog https://blog.keras.io/building-powerful-image-classification-models-using-very-little-data....
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 ...
18
votes
1answer
16k views

back propagation in CNN

I have the following CNN: I start with an input image of size 5x5 Then I apply convolution using 2x2 kernel and stride = 1, that produces feature map of size 4x4. Then I apply 2x2 max-pooling with ...
3
votes
3answers
40 views

How does “ Sparsity of connections” in CNNs causes the network to have less parameters?

I am studying Andrew NG's lectures on Convolutional Neural Network and he had provided two reasons for CNNs having less parameters compared to Non-Convolutional networks . They are : Parameter ...
0
votes
1answer
40 views

What is the best way of combining audio and visual data to make predictions?

I am trying to predict the probability of a disease by using audio and images, the audio and the images do not come from the same source. I am thinking of combining the outputs (maybe average them) of ...
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 ...
0
votes
0answers
15 views

Configuration of CNN model for recognition of sequential data - Architecture of the top of the CNN - Parallel Layers

I am trying to configure a network for character recognition of sequential data like license plates. Now I would like to use the architecture which is noted in Table 3 in Deep Automatic Licence Plate ...
0
votes
3answers
74 views

Problem with overfitting

I make small CNN from scratch to classify barcodes. I have two classes: one for images with barcodes and second for all what isn't barcodes (items, animals, landscape, furniture, people). I got good ...
0
votes
0answers
28 views

How to deal with severe overfitting in a UNet Encoder/Decoder CNN in a task very similar to image translation?

I am trying to fit a UNet CNN to a task very similar to image to image translation. The input to the network is a binary matrix of size (64,256) and the output is of size (64,32). The columns ...
4
votes
1answer
12k views

Keras Conv1D for simple data target prediction

I am trying to use conv1D layer from Keras for predicting Species in iris dataset (which has 4 numeric features and one categorical target). Following is my code: ...
8
votes
4answers
3k views

Faster-RCNN how anchor work with slider in RPN layer?

I am trying to understand the whole Faster-RCNN, From https://www.quora.com/How-does-the-region-proposal-network-RPN-in-Faster-R-CNN-work Then a sliding window is run spatially on these feature ...
0
votes
0answers
8 views

Local RoI features in Faster/Mask R-CNN

I use Faster and Mask R-CNN for various problems, including training GANs. I'm particular interested in local features that Region Of Interest module in Faster R-CNN extracts from feature layer(s). ...
0
votes
0answers
6 views

Segmentation-free character recognition on an image: Multi-label, multi-class or sequential image classification problem?

I have some images which look like this one: They exist of 3 possible characters (A-C) and a length of 4. Now, I would like to run a neural network, which recognizes each character in the picture ...
3
votes
1answer
79 views

Does it make sense to train a convolutional neural network on lo-res, use on hi-res pictures?

this is my first machine learning project and actually also my first question here. I am a novice to machine learning with a background in theoretical physics. I want to use a CNN to detect scratches ...
1
vote
1answer
19 views

CNN model contains several images that are null

I'm using a deep CNN with the ReLU activation function. When visualizing the layers (each with 32 filters), several of the filtered images are zeros. I am trying to reason why this may be happening? ...
12
votes
3answers
13k views

What is the difference between Dilated Convolution and Deconvolution?

These two convolution operations are very common in deep learning right now. I read about dilated convolutional layer in this paper : WAVENET: A GENERATIVE MODEL FOR RAW AUDIO and De-convolution is ...
0
votes
0answers
16 views

Definitive Values in Confusion Matrix

I built a convolutional LSTM model for the classification of 4-image time series. I used n keras ConvLSTM layers, followed by a time-distributed flatten and a few dense layers, finalized by a dense ...
5
votes
3answers
136 views

Disparity between training and testing errors with deep learning: the bias-variance tradeoff and model selection

I am developing a convolutional neural network and have a dataset with 13,000 datapoints that is split 80%/10%/10% train/validation/test. In tuning the model architecture, I found the following, after ...
0
votes
1answer
10 views

Unstable results in test mode with fractional max pooling in PyTorch

I make some variants of ResNet, originally found in TorchVision, modify them, train them and so on. What I have found is that even in .eval() mode, even if I load state right before evaluation, I ...

1
2 3 4 5
10