All Questions
Tagged with neural-network computer-vision
27 questions with no upvoted or accepted answers
4
votes
0
answers
449
views
Spatial Transformer Networks and Data Augmentation
We are all familiar with the famous Deep Mind paper STN.
Upon implementation, such as here, did anyone still use input data augementation such as affine transformations?
There are used to make CNN ...
- 273
3
votes
0
answers
65
views
Information Extraction from image / text - approach?
I need assistance with a ML project I am currently trying to create.
I receive a lot of invoices from a lot of different suppliers - all in their own unique layout. I need to extract 3 key elements ...
- 185
3
votes
0
answers
983
views
How to calculate Average Precision for Image Segmentation?
If I've understood things correctly, when calculating AP for Object Detection (e.g. VOC, COCO etc) the procedure is:
collect up all the detected objects in your dataset
sort the detections by their ...
- 131
3
votes
0
answers
230
views
How to improve the neural art algorithm?
I have been playing around the algorithm with tensorflow in this paper.
I tried to convert a photo to a Chinese ink and wash painring, but I got some strange patterns in the output picture(those in ...
- 31
2
votes
3
answers
382
views
Can the same CNN architecture be used for different data sets?
I have a CNN architecture that works well on 32x32x3 images. Can I use that same architecture for a data set made up of 28x28x1 images? (Both data sets have 10 classes). If this is possible, what ...
- 21
2
votes
2
answers
4k
views
How to properly save and load an intermediate model in Keras?
I'm working with a model that involves 3 stages of 'nesting' of models in Keras.
Conceptually the first is a transfer learning CNN model, for example MobileNetV2. (Model 1) This is then wrapped by a ...
2
votes
0
answers
283
views
Maximum number of classes YOLO net can recognize on mobile
I'm trying to make a mobile app on image recognition(Computer Vision Application) . Does anyone know whether modern day smartphones have enough processing power/memory to recognize, say about 1 ...
2
votes
1
answer
188
views
Doing a fine tuning after a transfer learning
I read about fine tuning and transfer learning for CNNs and was wondering if we can do fine tuning after using transfer learning on the same CNN? If so, will this increase the performance of the model ...
- 387
1
vote
0
answers
25
views
Suggestions for labeling regression data to improve model accuracy
I'm working on a convolutional neural network that should predict up to 3 (x,y) coordinate pairs representing the waypoints of a concrete path, given an input image. This network will be used to help ...
- 11
1
vote
0
answers
22
views
Attention mechanism: Why apply multiple different transformations to obtain query, key, value
I have two questions about the structure of attention modules:
Since I work with imagery I will be talking about using convolutions on feature maps in order to obtain attention maps.
If we have a set ...
- 181
1
vote
0
answers
23
views
Machine learning model (neural network or SVM) for unequal feature matrices size
I have feature matrices obtained from visual bags of words model for various dictionary sizes. Example, Nx5, Nx10, …., Nx15000. Where N is the number of samples and 5, 10, …15000 are the visual ...
- 11
1
vote
1
answer
364
views
Why are axes-aligned bounding boxes used in object detection
I understand (I think) why in object detection, the result is a rectangle:
it is a simple shape that can be defined by 4 variables (2 pairs coords of opposite corners or 1 pair of coords + width and ...
- 205
1
vote
0
answers
65
views
Denoising Prior to Image Classification
From what I have read, Denoising during preprocessing for image classification tasks seems to be a bit controversial.
While on one hand it might improve classification accuracy, the computational ...
- 123
1
vote
0
answers
136
views
What should be the best CNN model for Feature extraction of images for Image-Image search engine using LSH?
I have images something like the below:
And like this:
I have a huge data around 20M or so am I want to apply de-duplication and improve the search for this ...
- 323
1
vote
0
answers
17
views
Can convolutional network learn structural properties of one feature w.r.t to other?
I'm going through the literature on pose-estimation ( DeeperCut, OpenPose, MultiPersonPosetrack).
I'm interested in knowing whether these networks/ generally a CNN can learn properties (geometrical) ...
1
vote
0
answers
869
views
Examples for multi-input Convolutional Neural Network
I want to create a multi inputs Convolutional Neural Network (cnn) that takes two inputs and produces one output of the inputs class by using Keras. I searched for resources that explain multi inputs ...
- 2,015
1
vote
0
answers
88
views
How to estimate Distance to Obstacles using Lidar's BEV(Bir's Eye view ) Representation?
I am using a implementation that use both camera and Lidar for 3D obstacle detection for self driving cars but I have no idea how to calculate the distance to the obstacles since the Lidar uses the ...
- 387
1
vote
1
answer
155
views
Shape extracting on 2D geometric data
Given a set of lines, is there a way to train to extract geometric shapes. For example, the picture on left has some blue lines (with red endpoints). How can I train to extract shapes like on the ...
- 11
1
vote
0
answers
639
views
Trajectory data mining and pattern recognition using ORB-SLAM and KNN-DTW
I've started working on a project about the trajectory data mining from videos - for example: snowboarding video from GoPro action camera.
This is the continuation of my previous experiment (MotionML)...
- 111
0
votes
0
answers
8
views
Should I interleave sin and cosine in sinusoidal positional encoding?
I'm trying to implement a sinusoidal positional encoding. I found two solutions that give different encodings. I am wondering if one of them is wrong or both are correct. I showcase visual figures of ...
- 101
0
votes
1
answer
26
views
What is the "fast version" of ZFNet referenced in SPPNet and Faster R-CNN papers?
I'm reading old papers:
SPPNet: Link
Faster R-CNN: Link
In both cases, the authors refer to a "fast version of Zeiler and Fergus (ZF) Net"; specifically:
In SPPNet:
ZF-5: this ...
0
votes
0
answers
36
views
Losing Information while resizing the image in Segmentation task using U-net
I'm using U-net architecture to build a segmentation task of image. During training I have image of size 256256 image. It works very well on the segmentation of same size 256256 or near to size 256*...
0
votes
1
answer
66
views
Classification: ClassA vs. "everything else"
I am trying to create a neural network for recognizing a particular object. Maybe I am approaching this task from the wrong side, but, in my mind, this task boils down to teaching the network to do a ...
- 121
0
votes
1
answer
1k
views
What is `Multi-scale` in Multiscale Convolutional Network?
I was reading an article on Deep Learning and came across this term called Multi-scale Neural Network. I fully understand the concepts of convolutional neural network but it is a bit difficult to ...
- 103
0
votes
1
answer
169
views
Cable angle measurement (rotation)
I need to detect the rotation of a cable (degree) in the x-axis with high precision [0.2 (or more) degree detection] from its original state.
Detailed description:
I have a cable that is set in its ...
- 61
0
votes
1
answer
240
views
the size of training data set in the context of computer vision
Generally speaking, for training a machine learning model, the size of training data set should be bigger than the number of predictors. For a neural network, or even a deep learning model, the number ...
- 157
-1
votes
3
answers
3k
views
why the sigmoid function will be 1 and 0 if we use a fully connected layer that produce a big enough positive(res negative )output
HI I am using a fully connected network that uses sigmoid if we feed a a big enough weights the sigmoid function will finally become 1 or 0 , is there any solution to avoid this ?
and will this lead ...
- 387