Data Science Stack Exchange
Data Science Stack Exchange

Questions tagged [semantic-segmentation]

48 questions
5
votes
2 answers

What is Deep supervision?

I'm interested in segmentation models for medical imaging purposes. When I looked at the state of the art, I fell on a paper on a new architecture, Unet++: UNet++: A Nested U-Net Architecture for Medical Image Segmentation from Zongwei Zhou, Md…
Nicolas
  • 244
  • 2
  • 7
4
votes
2 answers

Tool for annotation of images for semantic segmentation

I have been searching around for a software tool, that I can use for annotating images. More specifically I want to do annotation to be used for semantic segmentation, meaning I want to create masks. I want to be able to create training data for…
lo2
  • 41
  • 1
2
votes
0 answers

What is the difference in computational cost at inference time between object detection and semantic segmentation?

I am aware that YOLO (v1-5) is a real-time object detection model with moderately good overall prediction performance. I know that UNet and variants are efficient semantic segmentation models that are also fast and have good prediction…
JStrahl
  • 161
  • 4
2
votes
0 answers

Image segmentation with large class imbalance leads to zero precision/recall

I have a binary semantic image classification problem where only very small parts of the images are positive, most of it is negative. In the training data I have a positive rate of around 0.023, which is a large imbalance (factor 43). Now when I…
Martin Ueding
  • 121
  • 3
2
votes
1 answer

How do you train a semantic segmentation model to optimize for IoU rather than accuracy?

I am currently building a U-NET semantic segmentation model on Tensorflow Keras to classify pixels as belonging to or not belonging to a class. For this problem, I've isolated the masks for only one class (even if multiple mask classes exist)…
stevemeisterr
  • 21
  • 3
2
votes
1 answer

Why is my training loss not changing?

I'm trying to train a semantic segmentation model based on this architecture, using this one as a base. The base model uses about 10 ReLU activations, and when implemented according to the first paper, the number jumps up to 14. The input images are…
Ad Ve
  • 23
  • 6
1
vote
2 answers

High image segmentation metrics after training but poor results in prediction

I'm trying to build a model with Keras that predicts four classes of features from microscopy noisy images which cover about 10 - 30 % of the image. I'm using U-net because my dataset is small (150 images for training and 30 images for validation).…
Capdi
  • 111
  • 1
  • 5
1
vote
1 answer

Pixel labelled image is must for semantic segmentation using deep learning in matlab?

I have large set of CT images containing lung regions. But I don't have corrosponding pixel labelled images. So how I could do semantic segmentation using DL in matlab. Like a test program I created 6 pixel labelled images using image Labeler. But…
Silpa
  • 11
  • 1
1
vote
1 answer

Correct way of computing dice score for image segmentation?

In binary image segmentation, for given a set of images, it's true mask and predicted mask. How do you compute dice score? Should I compute the dice score for each image separately and then find mean across all images? Or compute the dice score for…
spb
  • 88
  • 7
1
vote
0 answers

Semantic segmentation to instance segmentation

Having a mask from semantic segmentation I want to split it to list of non-overlapping instances masks(like an output from instance segmentation). Do you know some fast algorithm, approach, tool or library to do this? I have found one python…
Brans Ds
  • 849
  • 1
  • 8
  • 18
1
vote
1 answer

For semantic sementation, why am I getting better loss values with binary cross entropy than dice coef?

I'm learning all related to data science and how to train U-Net to do semantic segmentation. I have a U-NET with this loss function: def dice_coef(y_true, y_pred): y_true_f = K.flatten(float(y_true)) y_pred_f = K.flatten(float(y_pred)) …
VansFannel
  • 279
  • 1
  • 11
1
vote
0 answers

How to approach coarse-grained semantic segmentation?

I’m looking at doing something like semantic segmentation of images but where I only have pretty coarse-grained labels - roughly, for each 32x32 patch, I know if the answer should be “yes”, “no” or “unknown”. I’d like to start with a pretrained net…
Alex I
  • 3,152
  • 1
  • 23
  • 27
1
vote
0 answers

Per Channel loss or Per Sample Loss

I am currently tackling a semantic segmentation problem where, for each sample, my goal is to segment two masks corresponding to two objects. Notably, object two is typically located inside object one, and its size is significantly smaller. I am…
Ahmed Mohamed
  • 251
  • 3
  • 9
1
vote
0 answers

How do transposed convolutions in CNNs reduce the channel dimensionality?

In CNNs, I understand how convolution works and how it gradually reduces spatial resolution but increases the channel dimension. E.g. an RGB image of 100x100x3 after a few convolution layers may result in a layer that is 15x15x512 (i.e. 15x15 with…
tandeg
  • 11
  • 1
1
vote
1 answer

Building a CNN (with Keras for pixelwise classification)

I have a set of 120x120 input images with 3 channels. I want to build a basic CNN to predict the value of each pixel. I have 2 doubts. One is regarding the last layer - should be a Dense layer, or a Conv2D? model_CNN4 = keras.models.Sequential([ …
Filippo Nunes
  • 11
  • 1
1
2 3 4