Skip to main content

Questions tagged [convolutional-neural-network]

A convolutional neural network is a form of neural network with an additional convolutional layer, typically used in image & audio analysis. The convolutional layer is essentially a filtering stage defined by the kernel which is used. For example, a convolutional layer could have a kernel which extracts edges from an image towards the goal of learning which objects are in a scene.

Filter by
Sorted by
Tagged with
0 votes
1 answer
446 views

Do grouped convolutions actually improve learning?

My Understanding of Grouped Convolutions Let say we have some data with the dimensions [100,100,32] (lets ignore batch size and assume channels last) and we want to ...
J Houseman's user avatar
1 vote
1 answer
206 views

Does a rotational convolutional filter exist in neural networks?

Traditionally, a convolutional filter is one where you take a matrix of numbers, multiply it with a subset of the data, and then sum it up. Then you move the filter left to right and top to bottom in ...
xiaodai's user avatar
  • 630
1 vote
1 answer
486 views

How does the character convolution work in ELMo?

When I read the original ELMo paper (https://arxiv.org/pdf/1802.05365.pdf), I'm stumped by the following line: The context insensitive type representation uses 2048 character n-gram convolutional ...
Kaare's user avatar
  • 267
0 votes
0 answers
401 views

1D CNN time series classiifcation : ValueError: Shapes (10, 10, 8) and (10, 8) are incompatible

I'm working on a time series classification using ASHRAE RP-1043 chiller dataset which has 65 columns and more than 3000 rows for each chiller fault and normal condition. And I have used 1D CNN and ...
SINFER's user avatar
  • 1
0 votes
1 answer
720 views

Siamese Network for face comparison wont learn, accuracy stuck on 0.5, and loss stuck too

I'm trying to train a siamese network which contains a CNN and an embedding layer at the end to yield 2 similar (close) vectors for 2 images of the same person. I'm using the LFW_Cropped dataset, and ...
Jhon Margalit's user avatar
1 vote
0 answers
74 views

Specify torchvision transforms depending on the properties of an image and a mask

I have a dataset 1000 of images and corresponding segmentation masks from dermatologists. The images come in different sizes (as low as 400x600 and as large as 4Kx4K). 95% of image pixels are not ...
sixtytrees's user avatar
2 votes
1 answer
831 views

Why first fully connected layer requires flattening in cnn?

One can read everywhere on internet or in books that in convoluted neural networks, between convolution layers and the first fully connected layer, you should flatten your data. I managed to ...
chalbiophysics's user avatar
4 votes
1 answer
363 views

Can you use fully convolutional networks for binary classification?

I know that fully convolutional networks can be used for image segmentation and similar but I wondered if you could also apply them to simple image classification tasks. And if so, what is the proper ...
c0mr4t's user avatar
  • 173
2 votes
0 answers
36 views

Regress values inside the bounding boxes to predict a value in Object Detection

I am currently working on an object detection task. I have a dataset of grayscale and depth images. The annotation format is $x_1, y_1, x_2, y_2, class, depth$. I calculated this depth (of each ...
duddal's user avatar
  • 41
2 votes
1 answer
78 views

Studying and choosing between different neural network structures

I would like to develop a model that uses convolutional neural networks for image classification. From the many different network structures described in papers and articles online, I would like to ...
Curious Capybara's user avatar
2 votes
0 answers
125 views

Convolutional autoencoder - why keras example is asymmetry model?

I'm looking on keras convolutional autoencoder example, and confused with the model structure: ...
user3668129's user avatar
0 votes
2 answers
2k views

ValueError: Input 0 of layer sequential_7 is incompatible with the layer

I have 77 columns, with 4 class labels (already one-hot-encoded) by get_dummies. ...
Mimi's user avatar
  • 45
2 votes
1 answer
45 views

Can landmark detection be only used for faces and human bodies?

I want to use landmark detection for finding specific points of interest in an indoor setting e.g. bedrooms, bathrooms etc. Is it possible to use it? So far I have only seen landmark detection being ...
Yash Jain's user avatar
1 vote
2 answers
257 views

Average loss is 0 when training dataset with darknet yolov4

I am currently training a dataset using yolov4 darknet from AlexeyAB Github found here: https://github.com/AlexeyAB/darknet The dataset I am training is called FishNet Open Images. The dataset has 86,...
Dark Apostle's user avatar
1 vote
1 answer
44 views

Transfer Learning or Custom Network?

I am learning Computer Vision and I was wondering if it's usually worth it to build a custom convolutional network from scratch (through trials and errors) or if using transfer learning with a popular ...
Eric Cartman's user avatar
2 votes
2 answers
696 views

CNN + LSTM model for images performs poorly on validation data set

My training and loss curves look like below and yes, similar graphs have received comments like "Classic overfitting" and I get it. My model looks like below, ...
Vasanth Nag K V's user avatar
1 vote
1 answer
474 views

What is the intuition behind transposed conv layers being able to upscale images?

I was reading the ZF Net paper and it used the term Deconvnet on some searching it seems this is the wrong term and rather we use transposed convolutions instead. I understood how transposed ...
FoundABetterName's user avatar
1 vote
0 answers
71 views

What should be the input shape for convLSTM if ResNet-50 is applied before?

I have a video dataset, extracted all its frames, and applied ResNet-50 to extract features from all frames. ResNet-50 provides feature map of (2534, 7, 7, 2048), 2534 are the number of frames. Now I ...
TariqS's user avatar
  • 21
0 votes
1 answer
40 views

Why trainable parameters are not considered right?

I have tested the "ResNet" block and it works fine, but when I call it in the model class, it somehow it does not work properly? Is it related to the model definition?
Olfa2's user avatar
  • 1
1 vote
0 answers
31 views

What model to use for relative comparison between 3 figures?

I am working on a problem where I am given three images of different dishes (A,B,C) and the task is to figure out if figure B or ...
Buskruid's user avatar
1 vote
0 answers
41 views

Backpropagation on a CNN

I have tried searching for this, but it seems like no one addresses a key aspect of this problem (or maybe I'm overthinking this): So, first let's assume we have a 3x3 image with a single channel, and ...
Alex RD's user avatar
  • 11
1 vote
2 answers
3k views

High image segmentation metrics after training but poor results in prediction

I'm trying to build a model with Keras that predicts four classes of features from microscopy noisy images which cover about 10 - 30 % of the image. I'm using U-net because my dataset is small (150 ...
Capdi's user avatar
  • 111
1 vote
1 answer
365 views

Non Linearity used in LeNet 5

I was looking at the original implementation of LeNet-5 and I noticed a disparity in different sources. Wikipedia suggests that the non linearity used is the same sigmoid in each layer, some blog ...
FoundABetterName's user avatar
0 votes
0 answers
42 views

Weighting the loss function based on previous seen true positive rates

Similiar to class imbalance there is always something I would call "learnability imbalance" in multi-class classification. What I mean by that: Even when the classes are evenly distributed ...
T Piper's user avatar
  • 123
1 vote
2 answers
593 views

One hot vector output in classification task

I'm working on CNN model and I used one hot vector type of labels. The number of classes is 3: [1,0,0], [0,1,0], [0,0,1]. net(x) I'm getting such an output: [0....
Adolf Miszka's user avatar
1 vote
1 answer
21 views

Checking trained CNN on the images

I trained my CNN (model) classifier and want to check it on some new images. I have image x, so this syntax works for me for one image: torch.argmax(model(x)) What ...
Adolf Miszka's user avatar
1 vote
1 answer
403 views

Keras: apply multiple filters to each feature map in CNN

I am new to Keras, and I want to do the following: take a 2D image, and apply four 2D convolution kernels to it, giving four 2D feature maps. I could accomplish this. But then I want to apply two ...
user3433489's user avatar
0 votes
1 answer
1k views

Why is the kernel of a Convolutional layer a 4D-tensor and not a 3D one?

I am doing my final degree project on Convolutional Networks and trying to understand the explanation shown in Deep Learning book by Ian Goodfellow et al. When defining convolution for 2D images, the ...
puradrogasincortar's user avatar
0 votes
2 answers
3k views

Loading pretrained model with Pytorch

I saved my model with this code: from google.colab import files torch.save(net, 'model.pth') # download checkpoint file files.download('model.pth') Then uploaded ...
Adolf Miszka's user avatar
0 votes
0 answers
60 views

Testing trained model on the image from the test set

I trained my EfficientNet (CNN) and got accuracy=0.73. The question is how to check it on one concrete image from the testing set? How to write a code in python for it? I described the testing set ...
Adolf Miszka's user avatar
0 votes
1 answer
45 views

Adding Validation PyTorch

First of all, I'm new in this field and it's my first this kind of work. I'm trying to train EfficientNet (CNN), the code below is working fine, but I can't succeed to add also validation set to the ...
Adolf Miszka's user avatar
1 vote
0 answers
394 views

ROC and AUC curve for CNN multi-class classification problem

I have produced a convolutional neural network to classify images (malware images) into different classes/families. I have managed to produce a confusion matrix and classification report. My ...
Jack's user avatar
  • 155
3 votes
1 answer
36k views

ValueError: Classification metrics can't handle a mix of multilabel-indicator and multiclass targets in CNN

I am fairly new to ML and CNN, and this is my first attempt. I have managed to get my model to run, and now I am trying to produce a confusion matrix and classification report, but I am receiving an ...
Jack's user avatar
  • 155
0 votes
0 answers
78 views

How to interpret stagnant validation curve

I'm new to deep learning, so I'm just learning how to interpret my models. I'm creating a mixed-convolutional neural net to classify melanoma images. Here's the model structure: ...
Yehuda's user avatar
  • 113
0 votes
2 answers
1k views

Problems with shape of Conv1D on Keras

I have some problems with layers construction on Keras. I explain the whole problem: I have a feature matrix, with dimensions: 2023 (rows) x 65 (features); I tried to build a CNN, with Conv1D as ...
Martina Mamone's user avatar
0 votes
1 answer
35 views

Forecasting with Neural network and understanding which underlying model is favoured

If I have a very large set of data (~ 1TB). How can I use Neural Network on this data to understand which underlying distribution (eg. let's say a Gaussian or a Poissonian with a certain mean, sd) is ...
Ayan Mitra's user avatar
1 vote
1 answer
499 views

CNN can't predict images outside the dataset

I am using celeba dataset to train my CNN face landmark detection model. Here is my model ...
LOLs's user avatar
  • 21
0 votes
1 answer
145 views

Can the performance of a CNN be dependent on the train-test-val split random seed?

I am doing multi-class classification and comparing the effects of 2 image enhancement techniques (IET). IET 1 performs better than IET 2 at random seed x (for train-test-val split) IET 2 performs ...
djbacs's user avatar
  • 13
0 votes
1 answer
313 views

Does not using more filters in deeper CNN creates more images?

For example, we have applied 32 filters to a single image. Then it created 32 different images (stack of convolutional values). And in the second layer, if we apply 64 filters, are all these filters ...
EMT's user avatar
  • 113
4 votes
1 answer
315 views

why transpose convolution is called "transpose"

https://d2l.ai/chapter_computer-vision/transposed-conv.html https://en.wikipedia.org/wiki/Transpose I understand what transpose convolution does, but I am confused about the name of 'transpose'. In ...
richard.g's user avatar
  • 141
2 votes
1 answer
300 views

Training a neural network with TWO possible correct outputs for one input

I have a system as a black box that has two correct outputs for a single input sample. now I want to train a neural network to generate at least one of the correct outputs for that input sample. what ...
Abolfazl Sajady's user avatar
1 vote
1 answer
505 views

Correct way of computing dice score for image segmentation?

In binary image segmentation, for given a set of images, it's true mask and predicted mask. How do you compute dice score? Should I compute the dice score for each image separately and then find mean ...
spb's user avatar
  • 88
0 votes
1 answer
837 views

Testing accuracy very low, while training and validation accuracy ~ 85%

I have a training dataset of 10000 pictures and a test dataset of 15000 pictures. There are 23 types of birds. First of all, I imported the necessary ...
Trixiew's user avatar
0 votes
1 answer
471 views

How to calculate convolution for 2nd conv Layer in CNN, Do we need to average across all feature maps?

I understand that for the first layer (assuming we have a grayscale image) we calculate the convolution of 3*3 receptive field as a weighted sum of receptive weights with pixels $ x1 · w1 + x2 · w2 + ...
A.B's user avatar
  • 336
4 votes
1 answer
416 views

Using softmax for multilabel classification (as per Facebook paper)

I came across this paper by some Facebook researchers where they found that using a softmax and CE loss function during training led to improved results over sigmoid + BCE. They do this by changing ...
Steve Ahlswede's user avatar
0 votes
0 answers
37 views

Getting constant accuracies for training and validation sets despite their losses are changing during CNN training?

As the title clearly describes the issue I've been experiencing during the training of my CNN model, the accuracies of training and validation sets are constant despite the losses of them are changing....
talha06's user avatar
  • 101
0 votes
1 answer
494 views

Training loss = 0, training accuracy =1, validation and test around 85%

I have created different CNNs for doing image classification. The dataset is this: https://www.kaggle.com/crowww/a-large-scale-fish-dataset There are 9 classes, and each class contains 1000 images of ...
CasellaJr's user avatar
  • 229
0 votes
1 answer
495 views

Max Pooling in first Layer of CNN

I am seeing, in all the notebooks that I found, that Max Pooling is never used in the first layer of a CNN. Why this? Is it a convention among data scientist to do not use max pooling in the first ...
CasellaJr's user avatar
  • 229
2 votes
1 answer
743 views

Improve Convolutional Autoencoder

I just built a Convolutional Autoencoder to try to reconstruct a time series with shape (4000, 10, 30). This is the code, I used a batch size of 32, but I think it ...
Fabio's user avatar
  • 53
1 vote
1 answer
562 views

3 images as one input in CNN (U-Net) [closed]

I have been advised by my supervisor that if my U-Net segmentation network has RGB images at the input then I could use the channels for different images - median filter for R, normalization for G, ...
Tyler Durden's user avatar

1
3 4
5
6 7
15