6
votes
Accepted
Why do we need to concatenate in a U-Net?
These types of connections between layers are called skip connections, searching for this will give you a way to find more in-depth information. Broadly speaking, there are two advantages to using ...
- 8,197
3
votes
Accepted
What are features in computer vision?
CNNs like U-Net extract lower level features like edges on lower layers (i.e. the first convolutional layers) and higher level features on higher layers (i.e. convolutional layers closer to the final ...
- 5,490
3
votes
How can I create .nii (nifti) file from 3D Numpy array
You can this using nibabel:
import nibabel as nb
ni_img = nib.Nifti1Image(numpy_array, affine=np.eye(4))
nib.save(ni_img, "dicom_volume_image.nii")
- 31
2
votes
Modifying U-Net implementation for smaller image size
It appears that the original images are 68x68 pixels and the model expects 256x256.
You can use the Keras image processing API, in particular the smart_resize function to transform the images to ...
- 22.3k
2
votes
Algorithms to do a CTRL+F (find object) on an image
Invariant object recognition(IOR), refers to rapid and accurate recognition of objects in the presence of variations such as size, rotation and position.
SIFT and SURF are the most popular among them, ...
- 1,259
2
votes
Determining which deep learning model architecture is better
Not sure what language/framework you are using for your experiments but, you should try to deep seed your models so that performance is decoupled from random as much as possible. See an example here: ...
- 2,078
2
votes
Splitting a pdf containing batch of scanned documents
"Machine Learning for Digital Document Processing: from Layout Analysis to Metadata Extraction" by Esposito, Ferilli, Basile, and Mauro goes into detail about how to create a custom system for parsing ...
- 22.3k
1
vote
Google Earth Pro Satellite image segmentation using clustering
Whoa, are you trying to do clustering based on satellite images?! Just use the underlying long & lat coordinates to do this. It will be infinitely easier and so much more accurate too. Look at ...
- 613
1
vote
Accepted
How does the "skip" method work for upsampling? (fully convolutional NN)
So as you already know, the convolutions extract some local information from the image, but for some decisions you need more global info. This is what pooling is supposed to do, it basically erases ...
- 1,555
1
vote
multiple images inside one large CSV file
The workaround is you can place all the images on public drive accesible to everyone. In the csv file share the path of the file as one field with all other details. this should work.
You cannot load ...
- 2,074
1
vote
How to get the expected output shape from a unet model?
Your last layer uses 4 filter because num_classes is set to 4, resulting in an array with 4 channels in the last dimension. If you simply want only one channel ...
- 8,197
1
vote
Accepted
Yolov5 image detection without segmentation?
segmentation mainly uses Fully Convolutional Network(FCN) architecture. FCN is a CNN without fully connected layers(FC). segmenation can be thought as an encoder followed by a decoder. Here encoder ...
1
vote
resnet50 implementation for semantic segmentation
The output from the ResNet model is a vector containing the probability that the image belongs to each of the n classes, in your case to any of the 21 classes. If ...
- 8,197
1
vote
Accepted
Hook up PyTorch U-Net model to video
The RuntimeError error you're getting is caused by an incorrect number of dimensions in the input. The model expects an input with four dimensions with the first ...
- 8,197
1
vote
Accepted
Segmentation model to predict face forward and profile parts of the face
There is a library called facemesh for Python. It can detect face landmark points in Python. Connecting facemesh points to crop image to a desired polygon is ...
- 201
1
vote
Accepted
Implementing U-Net segmentation model without padding
$\hspace{3cm}$
If we follow the definition of each arrow.
Gray => Copy and Crop
Every step in the expansive path consists of an upsampling of the
feature map followed by a 2x2 convolution (“up-...
- 5,759
1
vote
Semantic segmentation in high-resolution images with high variance - cannot avoid underfitting
Given your mask is very small, you should look at reducing your convolutions to 2x2 since that will help aggregate more information from these smaller masks. EfficientNet has 3x3 and 5x5 convolutions ...
- 712
1
vote
Photorealistic synthetic data for object segmentation
I used such an approach at the Computer Vision Lab at TU Delft for my Bachelor thesis.
The goal was to analyze different aspects of the same problem: identifying LEGO bricks in a photograph of an ...
- 201
1
vote
Accepted
What is Deep learning approach to count the number of Diamonds in an image?
I say that go with Object Detection. Because you already have annoted images. And segmentation applied to locate objects and boundaries (lines, curves, etc.) in images. . In Recent days EfficientDet ...
- 281
1
vote
Accepted
What do the parameters used in crop mean?
Lets say my image is called frame and x, y, w and h are xmin, ymin, xmax and ymax
You're confusing $w$ with $xmax$ and $h$ with $ymax$: Usually $w$ is the width of the crop whereas $xmax$ is the ...
- 25.9k
1
vote
Tool for annotation of images for semantic segmentation
Diffgram is really great for this! I used it for a construction monitoring project. It's Open Source. From their site:
Semantic Segmentation Tools:
Auto Bordering: Automatically detects edges to ...
- 111
1
vote
Normalization of CT scans
There is extensive literature and established software to normalize CT scans. It is required for medical research to conduct group-level analysis. The most common process involves transforming to a ...
- 22.3k
1
vote
Extremely stochastic validation loss/accuracy
There can be a few reasons for this behavior as already pointed out in "Why is the validation accuracy fluctuating?":
Size of train / validation stes: Fluctuations may become stronger the ...
- 528
1
vote
Accepted
For semantic sementation, why am I getting better loss values with binary cross entropy than dice coef?
So the question asks about why different loss function lead to different error scores.
So globally error is there to help us measure the level of discrimination between the output of the model and the ...
- 1,468
1
vote
Accepted
Which colour channel from a TIFF image do I have to use?
From the readme of your kaggle link;
All images are provided in .tif format with 3 channels per image.
For 101 cases, 3 sequences are available, i.e. pre-contrast, FLAIR, post-contrast (in this order ...
- 350
1
vote
Accepted
Multiple output size in neural network
As you can see in lines 286-296 in newmodels.py the model can use two different loss functions for the four different outputs.
...
- 8,197
1
vote
Accepted
neural network probability output and loss function (example: dice loss)
I think there is a bit of confusion here. The dice coefficient is defined for binary classification. Softmax is used for multiclass classification.
Softmax and sigmoid are both interpreted as ...
- 1,555
1
vote
neural network probability output and loss function (example: dice loss)
The Dice coefficient tells you how well your model is performing when it comes to detecting boundaries with regards to your ground truth data. The loss is computed with 1 - Dice coefficient where the ...
- 785
1
vote
Accepted
Confusion matrix of UNET image sgemenation model
If your masks have only one channel and in the case of binary segmentation, you can easily compute a confusion matrix from one image thanks to:
TP: ground truth and predicted pixel are of class 1 (...
- 244
1
vote
How to prepare masks for multiclass semantic segmentation?
You should create a separate binary mask (1 for the pixlels belonging to that class and 0 for the rest of pixels) for each class. Therefore, your mask array should have a shape of (BATCH_SIZE, WIDTH, ...
- 111
Only top scored, non community-wiki answers of a minimum length are eligible