Artificial Intelligence Stack Exchange is @
question and answer site for people interested Anybody can ask a question x
in conceptual questions about life and
challenges in a world where "cognitive"
functions can be mimicked in purely digital Anybody can answer
environment. It only takes @ minute to sign up.
The best answers are voted up and
rise to the top
How should | incorporate numerical and categorical data as part of
the inputs to the U-net for semantic segmentation?
Asked 2 years, @ months ago Modified 10 months ago Viewed 312 times
‘a 1am using a U-Net to segment cancer cells in images of patients’ arms. | would lke to add patient data to it,
in order to see if it is possible to enhance the segmentation (patient data comes in the form of a table
0 containing features such as gender, age, etc). So far, my researches have led me nowhere, What can | do in
order to achieve this?
deep-learning data-preprocessing uenet_semanlie-segmentation — muli-tasklearning
Share Improve this question Follow cecited Mar 19,2021 at 11:50 asked Mar 18, 2021 at 1728,
bro
SS 394k 12 101 181
1 Answer Sorted by: Highest score (default) *
‘a. What you want to do is called multi-task learning. Here's what you do:
2 1 Cfeate second Input.
2. Attach it to 1D CNN (2-3 layers), so it aggregates this tabular information,
3. Concatenate this feature with the intermediate feature generated by the U-Net using Concatenate layer.
4, Puta dense layer of 2 after this.
A 5. Pat softmax with units = number of classes.
6. Add CE loss calculated with this one and the ground labels to the loss of U-Net.
‘This is in regards to TensorFlow. The same can be done in PyTorch easily.
answered Mar 18, 2021 at 1729