Movatterモバイル変換

Question 1

1vote

1answer

37views

What is the magnitude of a gradient jump to be considered an evidence of the exploding/vanishing issue?

I'm having some issues with the training of a convolutional neural network, as the loss initially decreases but suddenly it becames nan. I guess the problem could be related to some exploding/...

Luke83

13

askedOct 13 at 13:34

Question 2

0votes

0answers

18views

Preventing GPU memory leak due to a custom neural network layer

I am using the MixStyle methodology for domain adaptation and it involves using a custom layer which is inserted after every encoder stage. However, it is causing VRAM to grow linearly, which causes ...

Vedant Dalimkar

1

askedSep 29 at 16:00

Question 3

0votes

0answers

24views

Conceptual comparison of how attention works in (1) Plain RNN, (2) Models incorporating CNN, and (3) Transformers

I am learning Transformers and want to understand the roles of Self-attention and Multi-head attention, but first, I am interested in understanding how "attention" is used and how it applies ...

Big Tuna

1

askedSep 27 at 20:32

Question 4

0votes

1answer

58views

Model isn’t learning with 1D CNN

I relatively new to deep learning and I am working on a task where I tried to adopt a CNN architecture to do some prediction. My input data is of size (32, 3, 8192) and my target is of size (32, 1, ...

ty.fu

1

askedSep 13 at 14:17

Question 5

0votes

0answers

26views

neural network for learning the permeter of objects from binary images

I have a binary image that is the Canny output of an image containing some objects and my goal is to learn to recognise the corners of the objects and measure the perimeters. Example:The goal is to ...

Sameer Kulkarni

123

askedAug 26 at 6:09

Question 6

2votes

1answer

58views

Why is deskewing so critical for OCR models when CNNs can handle slight rotations in other vision tasks?

In many image classification tasks, convolutional neural networks can tolerate moderate rotations and still recognize the object correctly (e.g., a cat at +15° rotation is still classified as “cat”)....

dracule22

33

askedAug 14 at 9:16

Question 7

0votes

2answers

79views

What loss function to choose that will assign a higher penalty to false negatives than to false positives for regression task?

I am using a machine learning model to remove interference from range-doppler maps to detect targets. I am using a supervised approach, in which I give as input the range-doppler map of target+...

ThinkPad

43

askedAug 12 at 13:13

Question 8

0votes

0answers

36views

Neural Network for computing Holograms

I would like to build a neural network to compute hologram for an atomic experiment as they do in the following reference: https://arxiv.org/html/2401.06014v1 . First of all i dont have any experience ...

J.Doe

1

askedJul 28 at 8:39

Question 9

0votes

0answers

25views

Improving Segmentation on Messy image with very minimal labelled data?

I want to get the distribution of different lengths(biggest among width and height) of these chips from this type of image?My current idea is to use SAM 2.1 to generate segmentation of maximum no of ...

Garv Chaudha

11

askedJul 14 at 14:58

Question 10

0votes

0answers

21views

How to write temporal convolutional neural network?

I've heard temporal convolutional neural network is better for very long sequences than lstm and I've been dealing with just that and was seeing for potential architecture improvements (currently I've ...

AnArrayOfFunctions

121

askedJul 7 at 12:03

Question 11

1vote

1answer

51views

Best practices for integrating a single boolean feature in an image-based neural

I'm working on a binary classification task using a convolutional neural network (CNN). Alongside the image data, I also have access to a single boolean feature.I'm not an expert in feature ...

Antonio Rossi

111

askedJun 20 at 16:09

Question 12

0votes

1answer

41views

CNN Fixed Predictions

I’m building a Keras model based on MobileNetV2 for frame-level prediction of 6 human competencies. Each output head represents a competency and is a softmax over 100 classes (scores 0–99). The model ...

Bassel Attia

1

askedJun 9 at 8:43

Question 13

1vote

0answers

28views

Convolutional Kernels in CNN learning to find different patterns

Suppose we have an input image of dimensions $w \times h $ and the first hidden layer has dimension $(w-1) \times (h-1) \times 3$. We have $3$ seperate $3 \times 3$ kernels with no padding.I ...

Stan

11

askedApr 26 at 11:12

Question 14

4votes

1answer

218views

Are vision transformers scale invariant like CNNs?

I was trying to implement a vision transformer (RT-DETR) for object detection. I trained the model on 640x640 px images and tested it on a 2000x2000 px image containing many objects - the outputs did ...

Lockhart

143

askedApr 10 at 15:17

Question 15

0votes

0answers

54views

How to Improve Levenshtein Distance in CNN-BiLSTM Morse Decoder?

Problem Context:I'm building a Morse code audio decoder using CNN-BiLSTM with CTC loss. My current 4-layer model achieves Levenshtein distance ≈0.6, but attempts to improve performance by adding a ...

alexander

9

askedApr 9 at 19:35

Movatterモバイル変換

Stack Exchange Network

Questions tagged [convolutional-neural-networks]

What is the magnitude of a gradient jump to be considered an evidence of the exploding/vanishing issue?

Preventing GPU memory leak due to a custom neural network layer

Conceptual comparison of how attention works in (1) Plain RNN, (2) Models incorporating CNN, and (3) Transformers

Model isn’t learning with 1D CNN

neural network for learning the permeter of objects from binary images

Why is deskewing so critical for OCR models when CNNs can handle slight rotations in other vision tasks?

What loss function to choose that will assign a higher penalty to false negatives than to false positives for regression task?

Neural Network for computing Holograms

Improving Segmentation on Messy image with very minimal labelled data?

How to write temporal convolutional neural network?

Best practices for integrating a single boolean feature in an image-based neural

CNN Fixed Predictions

Convolutional Kernels in CNN learning to find different patterns

Are vision transformers scale invariant like CNNs?

How to Improve Levenshtein Distance in CNN-BiLSTM Morse Decoder?

Hot Network Questions

Subscribe to RSS