Questions tagged [computer-vision]
Questions related to image representation, segmentation, visual object categorization and image processing algorithms in general.
479
questions
0
votes
0
answers
9
views
How can we combine different Yolov8 models that each one has a different class [closed]
I have important question to continue my project and I’m seeking for help.
I trained 3 different Yolov8 model that each one detected specific object and has different dataset than other model. Now I ...
- 1
0
votes
0
answers
6
views
How to calculate IoU metric if there is no mask
For example: YOLOv8 segmentation doesn't output any mask for concrete class if no masks was detected, but there is a mask for this class in the ground truth masks.
I should set metric to zero for such ...
- 121
0
votes
0
answers
11
views
trying to create a cache dataset in monai: LinAlgError: SVD did not converge [closed]
This is the error that I am getting:
...
- 1
0
votes
0
answers
15
views
Searching for a proper way to reduce the dimensionality of activations from a CNN
I am conducting an analysis to compare the similarities between different images across early and late layers in a CNN. The model I am working with is the pretrained DenseNet121 that comes with ...
- 33
0
votes
0
answers
11
views
Instance Segmentation: Categorisation of architectures (one-stage / proposal based / etc.)
I am currently working on the topic of instance segmentation. So far I have found three possibilities for categorising the architectures:
one-stage / two-stage architectures
proposal based / proposal ...
0
votes
0
answers
22
views
Training an Image Captioning Model with variable number of captions per image
I am following this guide for training an Image Captioning model
It uses a dataset which always has 5 captions per image. My dataset greatly varies how many captions I have per image from 1-42. This ...
- 133
0
votes
0
answers
30
views
How to improve hough transform for line detection?
I'm looking for a method to detect false observations inside the hough transformation space, even called line detection algorithm.
The hough space can be described as this
X-axis is radius and Y-axis ...
- 381
0
votes
0
answers
11
views
Question about Bounding Box Regression in R-CNN
I am curious why, in that equation of Bounding Box Regression, the Feature Vector φ5 obtained from Pool 5 among the trained CNN Layers is multiplied by w. Also, I'm curious about the reason for adding ...
1
vote
0
answers
10
views
Any image datasets with Inter-annotator agreement (IAA) values recorded? [closed]
Is anyone familiar with publicly accessible image datasets that include Inter-annotator agreement (IAA) scores? Ideally for object detection or classification tasks.
Thank you!
- 11
0
votes
0
answers
44
views
Understanding relation between axis of least and maximum second moment
I was going through computer vision lecture video. You can find the pdf of this lecture here.
I was trying to understand how orientation of object corresponds to axis of least second moment aka ...
- 215
0
votes
0
answers
76
views
Very low train accuracy using ResNet and Efficientnet transfer learning medical image classification
I implemented and tested DenseNet, ResNet18, ResNet50 and Efficientnet from pretrained models in pytorch torchvision. Only denseNet121 is working. The training and validation accuracy are both very ...
1
vote
0
answers
9
views
Image segmentation service [closed]
hope you are well.
Does any one knows any service where I can send an image an return its segmentations areas, like using Meta's SAM (https://segment-anything.com/ )?
If not, does any one knows a ...
- 11
-1
votes
1
answer
75
views
Machine learning and Artificial intelligence algorithms in identifying and classifying Airplane parts
Airplane parts and functions
Can Machine Learning and Artificial intelligence Algorithms assist in identifying and classifying Airplane images parts?
- 119
0
votes
0
answers
25
views
Is Measuring Feature Similarity Using Cosine a Valid Method?
It seems to be a common method in deep learning papers to measure the similarity of features coming from the backbone using cosine similarity.
However, in my opinion, since in most cases, the features ...
- 1
1
vote
0
answers
17
views
How to measure effective sample size for multivariate time series?
I am working with images collected over time from a fixed camera. Since the scene doesn't change for long periods, my effective sample size is much lower than the actual number of images I've ...
- 641
1
vote
1
answer
58
views
Why does ViTForMaskedImageModeling not construct the original image correctly?
I was trying to use masked image modeling in huggingface and I saw ViTForMaskedImageModeling in their documentation but I did not understand how it reconstructs the original image ...
- 1,360
1
vote
0
answers
33
views
What would be the best way to find N clusters equally separated, ignoring outliers?
I am working on a computer vision problem (chessboard recognition), and my goal is to find the lines that correspond to the chessboard lines. I want to do this via Hough transform, since said lines ...
3
votes
2
answers
90
views
What are the best resources on image synthesis?
What are some good resources to learn about image synthesis? What are some of the key concepts or architectures to study?
I understand image synthesis as generating new images with ML techniques.
- 443
0
votes
0
answers
21
views
Looking for resources for supervised learning on video classification
I have quite a generic problem. I have a collection of videos, and a collection of tags that identify an action on a specific timestamp. I wish to be able to classify the correct neighborhoods of such ...
- 326
1
vote
0
answers
16
views
Active learning for Computer Vision where I have to generate my own images
I'm interested in applying active learning to a computer vision bounding box prediction project, where I don't have a large corpus of unlabeled images available, and instead have to take all pictures ...
- 11
0
votes
0
answers
13
views
Research Problem about Weakly Supervised Learning for CT Image Semantic Segmentation
My issue is that Grad-CAM often highlights the wrong areas. My task is to perform weakly supervised semantic segmentation (WSSS) of lung malignant tumors using image-level labels. Although I achieved ...
- 1
0
votes
0
answers
23
views
How can I evaluate the performance of a object detector at a fixed confidence threshold?
I have an object detector and now I have to decide which confidence threshold to use for each class. How can I determine what is the best confidence threshold for each class? Once decided, how can I ...
- 1
0
votes
0
answers
27
views
What is anchor-based and anchor free two detection paradigm in object detection?
I was reading this paper and saw this passage:
The architecture of these detectors has evolved from the initial two-stage [9, 26, 3] to one stage [19, 31, 1, 10, 22, 13, 36, 14, 7, 33, 11], and two ...
- 1,109
1
vote
0
answers
45
views
Questions on filtering out predicted boxes during training for YOLO family of algorithms
Does any of the YOLO models use NMS during training? From going through the papers they only explicitly state that NMS is only applied during inference, unlike Faster-RCNN.
Does any YOLO (or other ...
- 1,109
0
votes
0
answers
27
views
Which is more accurate, semantic segmentation or object detection?
In my project, which is product surface defect detection, such as to detect scratches on the washing machines, the target is to increase the precison and recall of detection.I think object detection ...
- 1
1
vote
1
answer
46
views
Continuous retraining a model on a increasing database
I have successfully developed an image classifier using Deep Learning, in particular I have used a ResNet50V2 network with fine tuning transfer learning. I built up a database from the available ...
- 51
0
votes
0
answers
4
views
implementing a new dataset on cv architecture
I want to implement the PIE dataset in the AgentFormer arch.
AgentFormer uses ETH and nuScene datasets. I successfully run these datasets on this arch. However, I couldn't take a good way with the PIE ...
- 33
0
votes
1
answer
39
views
what are the differences between receptive field (RF) and field-of-views (FOV) in DeepLab papers?
I am learning the deeplab models. However, some concepts in the papers made me confused. Receptive field (RF) and field-of-views (FOV) are two concepts mentioned in the Deeplabv1 paper. I know that ...
- 131
0
votes
0
answers
51
views
Implementing Patch-Based Segmentation for a Lung Lesion Segmentation task
Background:
I have two datasets, both with Lung CT Scan images as input. The labels for both are masks. The first dataset has Lungs as masks. The second dataset has lung lesions (17 classes like ...
- 41
2
votes
1
answer
54
views
What is the best approach for feature selection if I have 65536 features
Don't want to write a big paragraph and for the beginning, what I would like to say is I started doing a machine vision project and I flattened my 256x256 photos into a vectors of 65,536 features.
...
- 123
0
votes
0
answers
59
views
Huber loss problem for object localization
Object localization refers to finding a bounding box for an object in a frame for the purpose of object detection.
I have read here on page 2 that Huber loss has a problem in object localization as it ...
- 249
0
votes
0
answers
12
views
solving simple MRF problem with graph-cut
I'm trying to solve this simple question with graph cut:
I draw the graph and I get that that all pixel label sould be 1 while 000 is minimizing E,
what did I missed?
1
vote
0
answers
32
views
Neural Networks Miscalibration Measure
I have read these two papers related to the neural network miscalibration problem: "On Calibration of Modern Neural Networks" and "Multivariate Confidence Calibration for Object ...
- 11
6
votes
1
answer
1k
views
Which Deep Learning Method to use for the classification of precious stones (diamonds, sapphire, ruby etc) based on digital photo images and data set?
m trying to build Machine Learning Model for the classification of precious stones (like diamonds, sapphire, ruby) based on digital images. So with time the it was performed 150,000 gemstone ...
- 173
1
vote
1
answer
29
views
What are some best practices for labeling data that exists in a continuum?
I am building computer vision models on data that exists in a continuum. For example, imagine I'm trying to do semantic segmentation on cars. Some of the labels are distinct, like "chipped paint&...
- 408
0
votes
0
answers
67
views
Best practices in fine-tuning a model pre-trained with self-supervised learning for computer vision tasks
Self-supervised learning has been increasing in popularity recently in the computer vision domain as well. I was wondering if there are any practical best practices or tips and tricks one could follow ...
4
votes
1
answer
1k
views
How to determine if two images contain the same object without a dataset?
The problem I am trying to solve is, given two images, determining whether they contain the same object or not. Here is an example:
The first two images contain the same object, while the third image ...
- 141
1
vote
0
answers
54
views
Is it a good idea to have a category and its subcategories in the training set of an object segmentation model?
I hope you are doing great!
I am currently training an object segmentation model (detectron2 : mask rcnn)
The objective is to detect materials like wood, plastic, glass etc...
wood is one of the ...
4
votes
3
answers
2k
views
How to determine the no of multiplication operations in convolution operation?
Let's say We have an input of size 28×28×192. We apply 32, 5×5 convolution filters with padding "same". How many multiplication operations will be there in total?
I know there will be ...
0
votes
0
answers
34
views
Throwing up error on reconstruction of image data while training VQ-VAE
I have been working on the VQVAE model implementation based on the code(and formatting data) according to this link. This resulted in that I was able to run the model successfully and get the .pth ...
- 115
1
vote
1
answer
16
views
0
votes
0
answers
55
views
When to use Padding when Randomly Cropping Images in Deep Learning?
I am seeing these two options to process mini-imagenet images during training:
Option 1 torchmeta:
...
- 6,606
1
vote
1
answer
340
views
How is the variance for a diffusion kernel derived for a diffusion model?
So I'm watching this video tutorial from CVPR this year on diffusion models, and I am confused by the variance term in the distribution on the left on the video. I understand that in the forward ...
- 11
0
votes
1
answer
22
views
Are certain source models better suited for particular tasks?
I have the task of classifying medical images in a binary fashion. I plan on using transfer learning on a CNN but don't know what source model would be best to fine tune for this task.
Are certain ...
- 1
1
vote
0
answers
244
views
Does image data augmentation make sense when fine-tuning a transformer-based encoder-decoder model (Donut) on a small dataset (~100 samples)?
I am trying to fine-tune Donut model on ~100 (training) labelled data samples (pairs of images and json files). (Donut is a transformer-based encoder-decoder model, the encoder encodes images and the ...
- 250
1
vote
0
answers
129
views
YOLOv1 and YOLOv2 -- Pr(Object) and "responsible" boxes
In the YOLOv1 paper, Pr(Object) denotes the probability that there is an object in a certain grid cell.
When it comes to the ground truth values, which values can Pr(Object) take? Only 0 and 1?
What ...
- 35
2
votes
0
answers
463
views
α-balanced focal loss - why we actually decrease the importance of positive class
This is the equation for Focal Loss. The loss is an extension of weighted cross entropy, and aims to balance the impact of majority of easy negative class samples. The α parameter is a weighing term ...
- 686
2
votes
1
answer
90
views
How can fully convolution neural networks handle images of different sizes?
I've read that if we want to use images of different sizes in a convolutional neural network without resizing the images to a default size, we can use Fully Convolutional Neural Networks. But I do not ...
- 121
1
vote
1
answer
176
views
Use the Same Learning Rate to Train All Models When Doing Experiments For A Deep Learning Paper?
When writing a deep learning paper, I need to train several CNN models and compare their performances. They are from different architectures so different designs.
I'm wondering should I use the same ...
- 111
1
vote
0
answers
615
views
What is the correct way of normalizing/standardizing image-like data?
I have image-like data (e.g. H x W x C), where each channel contains quite different information. You can think of it being a 2D map (H x W) with information like elevation, wind velocity, temperature ...
- 11