Questions tagged [computer-vision]

Questions related to image representation, segmentation, visual object categorization and image processing algorithms in general.

Filter by
Sorted by
Tagged with
0 votes
0 answers
10 views

What is the purpose of zero mean in image processing? [closed]

I know that through zero mean and unit variance we normalize the image. But why is it good to have zero mean? What is the purpose of doing this? I have no transformation to do. I start with an image ...
user avatar
1 vote
0 answers
30 views

YOLO v2 loss function

I'm trying to understand (and implement) the YOLOv2 loss function, which is not given explicitly in the original paper. There are several posts on this topic, but quite a few seem to confuse the ...
user avatar
  • 11
0 votes
0 answers
8 views

How to fix this ValueError: Shapes (None, None) and (None, 3, 3, 16) are incompatible in VGG16 [closed]

I am currently fine-tuning a VGG16 on a multi-classification problem. The requirement is to add a new 1 Conv block, 1 Maxpool layer, 2 FC layers, and an output layer. I have removed the top layers of ...
user avatar
0 votes
0 answers
4 views

Resizing images for object detection using Faster R-CNN

I would like to detect several largest objects and extract features from images using Faster R-CNN. Given that Faster R-CNN was trained on images with the shorter side of 600 pixels. Does it make ...
user avatar
  • 2,401
0 votes
0 answers
9 views

Using Inception and FID scores in training?

Is it possible to use the Inception and FID scores in the training of a deep image generation model, i.e. to maximize the scores in a loss function, albeit this is "cheating"? If so, has ...
user avatar
  • 1
1 vote
0 answers
18 views

Balancing Multiple Evaluation Metrics for a Model

When evaluating a machine learning (or other statistical model) against multiple evaluation metrics, is there a standardized way to choose the "best" model? As a concrete example, for a two ...
user avatar
0 votes
0 answers
17 views

Best way to approximate head point having only face keypoints

I'm using the BlazeFace model from TensorFlow which only has this few keypoints: I need those keypoints plus a head keypoint, like this one: My question is, which would be the best way to ...
user avatar
0 votes
0 answers
11 views

Is there a dataset of images with varying sizes? [closed]

I'm working on a project dealing with image classification where I deal with images with varying sizes. I would like to validate my approach in other datasets. I would like to find other datasets with ...
user avatar
  • 378
0 votes
0 answers
3 views

Structure from motion approach to estimate target frame from nearby frames

My question comes from the paper: https://arxiv.org/pdf/1704.07813.pdf , which is an unsupervised learning approach for depth estimation. Suppose I have a sequence of images $I_{t-1}, I_t, I_{t+1}$. I ...
user avatar
  • 1
0 votes
0 answers
30 views

Is good to use binary images for deep learning?

I am trying to make a solar event detector from spectrograms. My problem is, I don't know how to approach the problem properly. I have generated high contrast images from the original data source in ...
user avatar
  • 1
1 vote
0 answers
8 views

Finding a dataset for a computer vision project related to medical imaging (related to cancer/tumor) [closed]

I am trying to find a dataset of medical images related to tumor/cancer, there should be images different stages of the cancer and also preferably the details about the patient, their medical history, ...
user avatar
0 votes
3 answers
38 views

How to calculate Precision and Recall for an image classification problem?

I'm not understanding how to calculate Precision and Recall if I'm doing image classification. If I have two classes, Cat and Dog, and for evaluation I get an image of a Dog and the model classifies ...
user avatar
0 votes
0 answers
8 views

Image-recognition model makes good predictions only with training examples

Im trying to use a kaggle dataset to train a model that recognizes american fingerspelling language from an image. The problem is that, built the model, if i record the screen with the examples ...
user avatar
0 votes
0 answers
6 views

Is classification with the backbone network a justifiable surrogate task in object detection?

There is a huge dataset of pictures but just a portion of the samples has bounding boxes. The main goal is to build an object detection model on the labeled dataset. One problem is that the dataset is ...
user avatar
1 vote
0 answers
25 views

What is the difference between these two types of training?

Suppose that I want to detect if a picture contains a particular logo, for instance the following one. Since template matching would be slow and fail those scaled or resized ones, I decided to train ...
user avatar
  • 5,151
0 votes
0 answers
10 views

Unconventional pretext task in computer vision - can I somehow justify it?

I was working on a industrial object detection neural network project. Since we had multiple images of the same object in different (but fixed) positions and light conditions, our dataset was very ...
user avatar
0 votes
0 answers
13 views

Extract Local Attention Positions

I am working on a project, I want to extract the local attention positions present in the feature maps of every level in the FPN. Below is the visualization: I am using MMDetection toolbox for ...
user avatar
0 votes
0 answers
33 views

Math behind predicting the dimensions of bounding boxes

I have been trying to understand the mechanics behind YOLO v3. I got stuck at the section which defines the calculation of bounding box. Going through various blogs, I found that The dimensions of ...
user avatar
  • 101
3 votes
1 answer
140 views

Adam is an adaptive learning rate method, why people decrease its learning rate manually?

Adam optimizer is an adoptive learning rate optimizer that is very popular for deep learning, especially in computer vision. I have seen some papers that after specific epochs, for example, 50 epochs, ...
user avatar
0 votes
1 answer
20 views

Classification with Augmentation vs Contrastive Learning

How Contrastive Learning based on (SimCLRv2 approach) is compared to regular classification (VGG, Resnset, etc) with Data Augmentation. It seems to me, that it should have very similar performance. I'...
user avatar
  • 583
0 votes
0 answers
8 views

How do you scale the activation function of an auto-encoder when using a custom normalization fitted on the data?

I'm working on a convolutional auto encoder. The input is an image The output is a reconstructed image During the training phase, we feed the same image in and out The loss is the Mean Squared Error ...
user avatar
  • 111
1 vote
0 answers
75 views

Multi-Head Attention in ViT

I need help to understand the multihead attention in ViT. Here's the code I found from GitHub: ...
user avatar
0 votes
0 answers
7 views

shape of output in feature pyramid network for RetinaNet

In FPN, each pyramid outputs a tensor which will go to classification and regression. This has an output of 256-d channels. But what is the complete shape of the output, including mini-batch size, ...
user avatar
1 vote
1 answer
37 views

Is Intersection over Union in object detection differentiable?

Is IoU in object detection differentiable or can be back propagated? Is it used in the training process or just in the inference?
user avatar
1 vote
0 answers
13 views

Annotations and Bounding Box Definition in Object Detection

I'm having confusion about the concepts of object detection. If the dataset has been annotated to have ground truth in (x, y) of the four vertices, then why are the papers (using the same dataset) are ...
user avatar
0 votes
0 answers
5 views

Can image tiling be used outside of remote sensing?

I am using computer vision (semantic segmentation)to digitise floor plans. I have read some papers which divide images in patches before the training.Once the training is finished, the patches are ...
user avatar
  • 1,019
1 vote
0 answers
33 views

Getting same prediction for all the classes in mobilenetV2 - Tensorflow

I am using mobile net v2 for multiclass image classification problem, here is how I am loading the data ...
user avatar
1 vote
1 answer
26 views

Question: Optimal D notation in Generative Adversarial Network (GANs)

I am completely new to Computer Vision and how Deep Neural Networks work on images in general. In particular, I have questions on the Discriminator component of Adversarial Generative Network (GANs). ...
user avatar
0 votes
0 answers
18 views

Can Mask-RCNN be used for semantic segmentation?

I am working on a computer vision project. I have learnt about semantic and instance segmentation and the algorithms: Unet, Mask-RCNN, etc. So far, after reading documentation I have associated Mask-...
user avatar
  • 1,019
2 votes
1 answer
37 views

How do anchor's play a part in the Region Proposal Network (RPN) in Faster-RCNN

I'm reading based on this article: https://tryolabs.com/blog/2018/01/18/faster-r-cnn-down-the-rabbit-hole-of-modern-object-detection as well as this YouTube video: https://www.youtube.com/watch?v=...
user avatar
1 vote
1 answer
40 views

Which algorithm or tool to use for object detection and extraction?

I will collect a dataset consisting of blackberries with and without defects. For this purpose, before I venture into this business I want to make sure that I can detect and extract blackberries from ...
user avatar
  • 23
1 vote
0 answers
22 views

Questions about mAP results from YOLOv2 paper

In the YOLOv2 paper, is the mAP metric displayed in Table 3 calculated in the same way as the metric in the column '0.5' from Table 5?
user avatar
  • 873
1 vote
0 answers
243 views

How are the confidence scores in YOLO computed?

I have read Yolo Loss function explanation but none of the answers discuss how box confidence scores are computed. The YOLO paper uses the following loss function: I'm confused about how confidence ...
user avatar
  • 873
1 vote
0 answers
16 views

Question about center coordinate definition in region proposal networks

In the YOLOv2 paper, page 3, it states that region proposal networks predicts $t_x$ and $t_y$ where $(x,y)$ center coordinates are calculated as: $$x=t_x*w_a-x_a$$ $$y=t_y*h_a-y_a$$ such that for $t_x ...
user avatar
  • 873
3 votes
1 answer
49 views

Question YOLOv2 bounding box prior

In the YOLO9000 paper, they define the distance between a box and centroid as $d(box, centroid) = 1 - IOU(box, centroid)$. I think box here is a ground truth bounding box, but what is centroid?
user avatar
  • 873
0 votes
0 answers
14 views

Multi-decoders may destroy VGG19 model property

A question about some statement that I didn't understood from the paper about stylization-based image colorization Stylization-Based Architecture for Fast Deep Exemplar Colorization (DOI: 10.1109/...
user avatar
1 vote
1 answer
39 views

Trainings Time of Original ResNet?

I am looking for Information on the original ResNet. I would be curious how long training roughly took. I know that a current challenge is to train the model as quickly as possible, but I really just ...
user avatar
  • 548
1 vote
0 answers
47 views

Online vs Offline Triplet Selection in FaceNet

I have been reading FaceNet. In the Triplet Selection section, it is written Generate triplets offline every n steps, using the most recent network checkpoint and computing the argmin and argmax on a ...
user avatar
1 vote
1 answer
182 views

When is Non-Max Suppression used in Object Detection

Is non-max suppression for bounding boxes obtained from a Region Proposal Network performed during training? From what I gather, NMS is not differentiable-- in which case, it can't be performed during ...
user avatar
1 vote
0 answers
27 views

how to check if unknown object partly covers the object i want to detect

I want to detect an object from an image,I use yolo object detection. I detect car from the image. I consider the case when another unknown object partly covers it, how can my program know that there ...
user avatar
0 votes
1 answer
153 views

Object detector evaluation (multi-class)

I have a (long) question regarding the evaluation of object detectors. I have two separate networks, the first one detects relevant objects (-> bounding boxes) and the second one classifies the ...
user avatar
1 vote
0 answers
43 views

Constructive learning models that can run (predict) fairly well on a Raspberry pi?

I'm looking for a model that is based on constructive learning and will do it's predictions live (via video camera) on the Raspberry Pi 4 Model B, either trained on a PC or pre-trained, I found out ...
user avatar
  • 111
3 votes
0 answers
29 views

Fisherfaces/Linear Discriminat Analysis - What are those faces supposed to be?

I see a lot of weird blue/gray/green faces when I search for "firsherfaces" at Google. I see faces like this and my question is simple: What are those faces supposed to be? Are they some ...
user avatar
  • 173
0 votes
0 answers
14 views

Is there any standard way of measuring the intra- and inter- class variance in image datasets?

I'm dealing with an image classification problem and I have a feeling that we have a small inter-class variance and a proportionally big intra-class variance. However, it would be interesting to use ...
user avatar
  • 378
1 vote
1 answer
76 views

Hand Keypoint Detection Model Not Converging

I'm currently trying to train a custom model with tensorflow to detect 17 landmarks/keypoints on each of 2 hands shown in an image (fingertips, first knuckles, bottom knuckles, wrist, and palm), for ...
user avatar
0 votes
0 answers
30 views

What is the reference point for predicting x,y coordinate in the bounding box regression (x,y,h,w) of the Faster RCNN model?

The Faster RCNN paper introduces a Region Proposal Network (RPN) that is trainable. This region proposal network outputs certain regions where objects can be located. Then we perform ROI pooling on ...
user avatar
0 votes
0 answers
81 views

connection between multi-label classification and multi-class classification

For a dataset with multi-label judgement, e.g., coco dataset but where we only want to predict the most-possible label. There're multiple ways, for example : 1) train as a multi-label learning(each ...
user avatar
  • 119
1 vote
0 answers
72 views

In YOLO, can there be more than one class within a cell?

I'm confused by the fact that YOLO** can have multiple bounding boxes per cell, but there is only one class vector. Does YOLO allow the truth/training to have more than one object in a cell, each with ...
user avatar
  • 509
0 votes
0 answers
21 views

Stacking/Blending for Convolutional Neural Networks using original training data

I have a dataset consisting of multiple classes of images. I trained several CNN's to predict the probability for those classes. Afterwards, I combined the predictions of the CNN's using a weighted ...
user avatar
  • 115
1 vote
0 answers
58 views

Does atlas-based imaging segmentation generally involve machine-learning [closed]

Segmentation is an important task in medical imaging analysis. Many FDA approved medical device use "atlas-based" segmentation tasks. Newer device use "deep-learning based" ...
user avatar
  • 80

1
2 3 4 5
9