issues
search
dlmbl
/
unet
0
stars
0
forks
source link
2024 recap
#5
Open
neptunes5thmoon
opened
2 months ago
neptunes5thmoon
commented
2 months ago
ran out of time (didn't make it to receptive field and shift equivariance)
would be nice to find a way to allow for groupnorm for semantic segmentation
bug in check valid function (fixed in dlmbl-unet version) and testing
students struggled with the num feature maps functions
IDEAS for 2025:
schedule more time
make tasks a little easier
add translational equivariance cropping
improve receptive field size plotting
double check receptive field size computation
scaffolding for computing feature maps might be too restrictive
IDEAS for 2025: