issues
search
lucamarini22
/
marine-anomaly-detection
Semantic segmentation of marine anomalies using semi-supervised learning (FixMatch for semantic segmentation) on Sentinel-2 multispectral images.
GNU General Public License v3.0
6
stars
0
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Is this published?
#40
robmarkcole
opened
7 months ago
1
L1c train
#39
lucamarini22
closed
1 year ago
0
Fixes
#38
lucamarini22
closed
1 year ago
0
nice-to-add colored output for logs
#37
lucamarini22
opened
1 year ago
1
Fixes
#36
lucamarini22
closed
1 year ago
0
Fixes
#35
lucamarini22
closed
1 year ago
0
Log val miou
#34
lucamarini22
closed
1 year ago
0
Channel importance
#33
lucamarini22
closed
1 year ago
0
Small fixes
#32
lucamarini22
closed
1 year ago
0
Softmax and augmentations fixes
#31
lucamarini22
closed
1 year ago
0
Grad lu fix
#30
lucamarini22
closed
1 year ago
0
Fixes
#29
lucamarini22
closed
1 year ago
0
One thing to think about in general maybe if you want your code at all to be usable with other datasets. Because it seems to me you are writing it in a way that will make it extremely difficult to ever use any other dataset? 🤔
#28
lucamarini22
opened
1 year ago
0
Turbid Water changes too, but yeah I need to make a script to get this numbers automatically when running training instead of getting that beforehand
#27
lucamarini22
opened
1 year ago
0
I would suggest you start using cfg files and just have a debug.cfg etc. Right now it seems you are changing default values in your code all the time. This is very dangerous as you may introduce errors or forget to change things back. What I would suggest is:
#26
lucamarini22
opened
1 year ago
0
Ssl one train set
#25
lucamarini22
closed
1 year ago
0
Ssl one train set
#24
lucamarini22
closed
1 year ago
0
Original classes baseline
#23
lucamarini22
closed
1 year ago
1
Implementation of Semi-Supervised Learning with 1 training set instead of 2 different training subsets
#22
lucamarini22
closed
1 year ago
6
Generation l1c update
#21
lucamarini22
closed
1 year ago
0
Hyperparameter tuning
#20
lucamarini22
closed
1 year ago
1
L1c update
#19
lucamarini22
closed
1 year ago
0
open todo
#18
lucamarini22
opened
1 year ago
0
Fixmatch semantic segmentation for multispectral data
#17
lucamarini22
closed
1 year ago
3
main to fixmatch-sem-seg
#16
lucamarini22
closed
1 year ago
0
Fixmatch sem seg
#15
lucamarini22
closed
1 year ago
0
If I was you, I would merge individual readme files into one big on the root. Otherwise, people will not find it or might discard your repo initially thinking that there are not enough docs etc.
#14
lucamarini22
opened
1 year ago
0
you are not dividing, not sure if intentional? if so, would remove to avoid later being confused whether you meant to divide
#13
lucamarini22
opened
1 year ago
0
I would move the constants bellow (class_distr) into a separate file. Or you can make there formatting nice with
#12
lucamarini22
closed
1 year ago
1
if you put your comments on the line above the code it may be less ugly.
#11
lucamarini22
closed
1 year ago
1
this is a very long file with multiple classes. would break up into individual files
#10
lucamarini22
closed
1 year ago
1
division by 0 for white / black img :P
#9
lucamarini22
closed
1 year ago
1
This is quite a large file with a lot of different functions some which don't have to be in this class since they are static. would consider moving some of out the class
#8
lucamarini22
opened
1 year ago
0
instead of having different arg parse in files may be nicer to have that collected in some other file? you probably need it in any your scripts?
#7
lucamarini22
opened
1 year ago
0
Some comments here and there. In general, I would try to break up your file structure a bit. If I read your filenames it is pretty hard to understand how things interact and in most files you have a lot of static functions, so it seems like a file is not something that combines things that have one clear purpose but just a bunch of thrown together functions. Would try to divide by purpose / activity into subfolders like : imageprocessing, io, datastructure or something like that with dedicated files for specific things
#6
lucamarini22
opened
1 year ago
0
open TODO
#5
lucamarini22
closed
1 year ago
1
open todo
#4
lucamarini22
closed
1 year ago
1
L1c
#3
lucamarini22
closed
1 year ago
2
add focal loss and update process of creating models and evaluating them
#2
lucamarini22
closed
1 year ago
0
Add 2 types of clustering of gt classes: binary and multi
#1
lucamarini22
closed
1 year ago
0