Questions tagged [cnn]
Convolutional Neural Networks (CNN, also called ConvNets) are a tool used for classification tasks and image recognition. The name giving first step is the extraction of features from the input data.
1,365
questions
1
vote
0
answers
39
views
Tensorflow model structure is strangely deformed in model.fit()
I was able to confirm (by checking model.summary()) that a model with the correct structure was successfully created. However, when ...
- 11
1
vote
0
answers
205
views
Batch size and steps per epoch
My data size is 6011, which is a prime number, and therefore, the only batch size number that divides this data evenly is either 1 or 6011. However, I need the batch size to be 32, which means that ...
- 145
0
votes
0
answers
21
views
Replicate the CNN arhitecture from a paper
In the paper it says:
We train our model using CNN with the number of filters 128, stride of
1, and kernel size of 3, 4, and 5. We then apply the ReLu activation
function with Max Pooling to the out ...
- 198
1
vote
1
answer
181
views
How does Alexnet flatten operation go from 6x6x256 tensor to 4096 vector
In the Alexnet model, after the encoder steps are completed, you end up with a 6x6x256 tensor. Now this needs to be flattened before we go to the ANN part of the ...
- 111
2
votes
1
answer
2k
views
Why is my training loss not changing?
I'm trying to train a semantic segmentation model based on this architecture, using this one as a base.
The base model uses about 10 ReLU activations, and when implemented according to the first paper,...
- 23
2
votes
0
answers
60
views
Which model is used for document extraction (CamScanner, Microsoft Lens etc)
I want to start a small project where I'd create a model(s) that would extract document from a picture and rescale it, something like CamScanner or Microsoft Lens apps do.
I've gathered a small ...
- 21
1
vote
0
answers
17
views
Looking for multi output image datasets
I'm looking for image datasets that have multiple labels. So far I could only find one dataset of age, sex and ethnicity prediction but I'm looking for something a little less known than that one. ...
- 11
2
votes
1
answer
4k
views
Does validation_split in tf.keras.preprocessing.image_dataset_from_directory result in Data Leakage?
For a binary image classification problem (CNN using tf.keras). My image data is separated into folders (train, validation, test) each with subfolders for two balanced classes. Borrowing code from ...
- 23
0
votes
0
answers
27
views
Augmentation for sound recognition of dog barks for CNNs
I am training CNNs to recognize dog barking, and for this I would like to augment the data sets I have (~30'000 10s clips with either barks, or no-barks in them).
The straight forward idea was to mix ...
1
vote
1
answer
42
views
spot/stain growth in image classification problems
I am working on a problem with images where we are monitoring development of spot in certain region of image. We are able to classify spot present(NOK) or not present(OK) successfully if initially ...
- 11
1
vote
1
answer
261
views
Is this Double U-Net overfitting?
I'm working on a undergraduate project with using deep learning. Currently, I'm trying to improve a model by modifying it. Model is Double U-Net and dataset that I'm using is DRIVE dataset. It ...
1
vote
1
answer
77
views
Merging Multidimensional Features in Python
I am trying to do 10-fold cross-validation on an audio dataset. The audio is clipped into small segments, so we have multiple clips from the same file. To avoid overfitting, each audio is assigned to ...
4
votes
4
answers
1k
views
Why are deep learning models unstable compare to machine learning models?
I would like to understand why deep learning models are so unstable. Suppose I use the same dataset to train a machine learning model multiple times (for example logistic regression) and a deep ...
- 151
1
vote
0
answers
21
views
If an FCN accept rectangular image as input or has to be square?
Some say that for FCN it doesn't matter if the input image is rectangular the only thing matters that the size must be constant ...
- 11
1
vote
1
answer
30
views
Question about performance of a CNN
I have a question (beginner :D) that is related to throughput and inference rate.
Can the throughput and inference rate change as the model is trained or are the values for these parameters fixed?
...
- 11
1
vote
1
answer
309
views
Loss is coming very high for object detection
I am using total of 3000 images for training an ssd_inception_v2_coco as the object detection model.
I have set batch size as 4 because I don't have a high end GPU hence I am renting it for few hours ...
- 11
10
votes
2
answers
3k
views
Why should I understand AI architectures?
Why should I understand what is happening deep down in some AI architecture?
For example LSTM-BERT- Partial Conv... Architectures like this. Why should I understand what is going on while I can find ...
- 117
0
votes
0
answers
155
views
Why do we use multiple convolutional layers, instead of a single layer
My intuition is that, when we have the raw pixels from the image, suppose we want to extract an eye, why don't we just try to use an eye detector to extract the eye feature from the image, why do we ...
-1
votes
1
answer
8k
views
ValueError:Input 0 of layer sequential is incompatible with the layer
i represented each row in my dateset as a 15552 cell which is spectrogram colored image(72723), that is represent the audio features, 72*72 is the size of spectrogram image and 3 referred to the 3 ...
- 37
4
votes
0
answers
235
views
ResNet50 + Transformer
In many papers people extract features from image using ResNet and than pass them through transformer. I want to implement the same. I want to get features and than classify them using transformer. ...
1
vote
0
answers
22
views
CNN-LSTM for price estimation
I have some pictures that I would like to extract features from using ResNet, and then apply another model like LSTM with extracted features to estimate the price.
I have made two folders, one ...
- 11
0
votes
1
answer
121
views
ResNet output dimensions of initial convolution don’t yield in an integer
I am trying to understand the ResNet dimensions, but got stuck at the first layer. We are passing a [224x224x3] image into 64 filters with kernel size 7x7 and stride=2. According to the ResNet source ...
- 3
0
votes
1
answer
53
views
Understanding Max Pooling
I understand max pooling in CNNs can help decrease the computational load due to downsampling.
Another thing mentioned is that max pooling can help provide a sort of "spatial invariance" for ...
- 1
0
votes
1
answer
1k
views
Should rescaling be used on test images in keras?
I am kind of confused regarding the topic. I have built a CNN architecture for the cat-dog image classification around 6000 images of cat and 6000 images of dog and I am predicting on test images. I ...
- 193
0
votes
0
answers
24
views
Does time-series dependence penalize prediction with CNNs?
I'm trying to use CNNs on time series data (EEG), measured on different people. Each person has 10-20 recorded signals of different lengths and every subject has one global class assigned.
Example:
...
- 1
1
vote
1
answer
312
views
How to load a dataset with a specific structure in tfds library?
I have a dataset that it's classes arranged in the following way:
/dataset/train/images/class1/
/dataset/train/images/class2/
.
.
.
/dataset/train/images/classN/
...
0
votes
1
answer
84
views
Improving the performance of neural networks
I have 3 questions in mind about the neural network
For the best model performance, is it better to train a model only on high resolution images or does it not matter whether the training data ...
- 25
1
vote
1
answer
1k
views
Keras ImageDataGenerator unable to find images
I'm trying to add image data to a Kaggle notebook so I can run a convolutional neural network but I'm having trouble doing this via ImageDataGenerator. This is the ...
- 135
0
votes
1
answer
69
views
Improving validation losses and accuracy for 3D CNN
I have used a 3D CNN architecture, for detecting the presence of a particular promoter (MGMT), by using FLAIR brain scans. (64 slices per patient). The output is supposed to be binary (0/1).
I have ...
- 103
1
vote
0
answers
236
views
Q: Training a CNN-LSTM on video inputs
Hello everyone!
I implemented the following model, for action classification from videos, where each frame is 224x224x3, a video consists of ...
- 11
0
votes
2
answers
2k
views
High accuracy in mode.fit but low precision and recall. Overfit? Unbalanced? Error?
Hello ive been training a CNN with keras. A binnary clasificator where it says if a depth image has a manhole or not. Ive labeled manually the datasets with 0 (no manhole) and 1(it has a manhole). I ...
0
votes
1
answer
392
views
How many layers of a pretrained model shoud be frozen?
I am following a transfer learning example where the blogger has frozen the first 20 layers of MobileNet. My question is, is there any rule of thumb for how many layers should be frozen? What is the ...
- 231
0
votes
0
answers
18
views
Differentiation of Learning Capabilities of Different Networks
I have a conceptual problem regarding the overall learning capability of a neural network differentiated by the different types of input that we can give to the network. Suppose that we have a ...
- 101
1
vote
3
answers
829
views
100% Accuracy and 0 loss in image classification
I am working on image classification using CNNs and the pretrained model VGG16, my dataset has 3 classes with almost 900 images per class. after traning for 5 epochs my model reached 1 accuracy with 0....
- 31
0
votes
1
answer
109
views
after overcoming the overfitting, how to increase training accuracy?
I am building a CNN using keras for a classification task. I started with a simple model as a starting point and as almost all ML problems go, especially if the dataset is not very big, I faced an ...
- 21
1
vote
0
answers
78
views
Identify areas within a shape/polygon with Vision / ML
Given a shape, in the format of a binary image, I would like to detect and subdivide it to new areas. Below is an attached example of such a shape and the expected outcome where each new area is ...
- 11
0
votes
1
answer
773
views
What does Keras image generators do with input images samplewise_std_normalization= True?
I have trained a a convolutional network samplewise_std_normalization=True. Now I want to check my model in real-time using Opencv. Therefore I would like to perform the same preprocessing on the ...
- 231
1
vote
0
answers
307
views
0
votes
1
answer
267
views
Reducing Validation loss for Triplet Loss Embeddings
I'm trying to create a facial recognition detector using triplet loss followed by a kNN algorithm.
I have roughly 10000 input images with 3 different classes, input size is 80x80. Model structure uses ...
- 101
0
votes
0
answers
1k
views
ValueError: `validation_split` is only supported for Tensors or NumPy arrays, found following input: RaggedTensor
I have following inputs to be trained on a CNN:
x = np.array(Images)
y = [ [[0]], [[76., 5., 9., 1., 0., 0.], [54., 4., 10., 51.]] ]
Since the ...
0
votes
1
answer
53
views
Tensorflow parameters for CNN
I created the below simple model (taken from a Coursera course). It has a total of five convolutions.
...
- 133
0
votes
1
answer
40
views
Image Preprocessing [closed]
I'm working on a use case where I need to pre process the image for my AIML model evaluation and I want to count all black pixels in RGB image.
Instead of iterating rows*column, I'm looking for some ...
- 1,252
0
votes
1
answer
124
views
How to custom conv2D layer Keras using calculated values
This is my first question, Hello World I guess.
I need to create a conv2D custom layer (at least, I think so), which should use my custom module for extracting values in the first layer. It would be ...
0
votes
2
answers
785
views
0
votes
1
answer
27
views
Test data accuracy from real world have lowest accuracy than validation data collected in simulation environment
Background:
Problem type: Multi class classification
The dataset contains around 1,000 samples (simulated dataset of sensor signals), where each sample is 2D i.e (1000 * 1000 * 8).
Additionally, I ...
- 165
1
vote
1
answer
104
views
Activation Function
I am very new to machine learning and made an experiment myself. I have a few questions:
Can I use $Y = sin(x)$ or $Y = 2x$ as an activation function for a neural network?
Is it necessary to increase ...
- 13
1
vote
0
answers
34
views
Model Performance is Fluctuating
I am training a 3D u-net model for 3D medical images. My training data has 800 images, the validation data has 200, and test data has 200 images. when I try to fit the model, there is a fluctuation in ...
- 36
0
votes
1
answer
31
views
Problem with CNN [closed]
I am using the BreakHis database. More specifically, I am trying to classify the 400X images. The sizes of the images are $700x460x3$.
Here are the details of the dataset. Also, here is the code for ...
- 221
0
votes
1
answer
1k
views
How to eliminate Non-Trainable params in Deep Learning [closed]
First of all, I would like to know what is the cause of Non-Trainable parameters?
Secondly, how do you eliminate them?
I used a combined CNN-RNN, it returned that 130 Non-Trainable parameters.
Thank ...
- 45
0
votes
1
answer
241
views
How to decide the padding size and stride size in CNN
In CNN in 2d, what situation is the size of the padding and stride changed in?
So far, I could make sense of the basic concepts with padding and stride.
Padding and stride can be used to adjust the ...
- 101