issues
search
shuaigezhu
/
star_galaxy_classification
0
stars
1
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
top k (like 50%) samples in loss function to train
#42
shuaigezhu
opened
4 years ago
0
0% for other pixels
#41
shuaigezhu
opened
4 years ago
0
To obtain more data
#40
shuaigezhu
opened
4 years ago
0
use SK-Learn to show Normalized Confusion Matrix
#39
shuaigezhu
closed
4 years ago
2
Increase the size of the validation and test set, then retrain the 6 models. Currently, there are 7700 in the training set, 270 in the validation and 270 in the test set.
#38
shuaigezhu
closed
4 years ago
1
Applied data augmentation to balance the number of quasar, star and galaxy
#37
shuaigezhu
opened
4 years ago
0
Applied data augmentation to balance the number star, quasar and galaxy.
#36
shuaigezhu
opened
4 years ago
0
Confusion Matrix
#35
kmyi
closed
4 years ago
1
Prepare Presentation
#34
kmyi
closed
4 years ago
2
Write a report using CVPR format (4--8 pages)
#33
kmyi
closed
4 years ago
0
Debug And Visualize the Gaussian Blur
#32
kmyi
closed
4 years ago
1
compare precision between MSE and KLDivloss
#31
shuaigezhu
closed
4 years ago
1
apply gaussian blur
#30
shuaigezhu
opened
4 years ago
10
Using MSELoss with different weights to replace KLDivloss
#29
shuaigezhu
closed
4 years ago
3
use Brier score to evaluate the prediction
#28
shuaigezhu
opened
4 years ago
0
implement Gaussian blob in label space.
#27
shuaigezhu
closed
4 years ago
1
Using Zscale to make images look clear.
#26
shuaigezhu
closed
4 years ago
0
masking the 'not a number' pixels for the input images.
#25
shuaigezhu
closed
4 years ago
3
Using the contour plot to visualize the prediction of classes
#24
shuaigezhu
opened
4 years ago
14
train Unet with 5 bands (g, r, i, z, u)
#23
shuaigezhu
closed
4 years ago
0
training Unet model with all images
#22
shuaigezhu
closed
4 years ago
0
save label image as one fits.gz file per input file
#21
shuaigezhu
closed
4 years ago
0
Spread up label pixel
#20
shuaigezhu
closed
4 years ago
0
change loss function to NLL with softamax
#19
shuaigezhu
closed
4 years ago
0
Cropping images as size of (256,256) based on objects.
#18
shuaigezhu
closed
4 years ago
0
increase the size of the training set then make a training set, validation set, and test set.
#17
shuaigezhu
closed
4 years ago
0
write a python script to download all the images in HSC footprint
#16
shuaigezhu
closed
4 years ago
0
add kqcg catalogue then make a new probability catalogue
#15
shuaigezhu
opened
4 years ago
0
compared the converted x and y (from ra and dec) to check out whether the converted x and y is right.
#14
shuaigezhu
closed
4 years ago
0
modify algorithm to update the probability catalogue
#13
shuaigezhu
opened
4 years ago
0
make a python script to allow cpu cluster to convert ra, dec to x, y on images
#12
shuaigezhu
opened
4 years ago
0
restrict objects from ra and dec.
#11
shuaigezhu
closed
4 years ago
0
Train UNet with 100 images
#10
kmyi
closed
4 years ago
2
Restrict objects in ra,dec space
#9
kmyi
opened
4 years ago
0
Write a simple Simbad query
#8
kmyi
opened
4 years ago
0
create probability image with 3 channels of prob_yufeng_{star,qso,gal} with projected ra,dec into pixel x,y, and save image as <orig_image_name>_prob_class.fits
#7
kmyi
closed
4 years ago
1
find matching images in pitcairn which includes the ra,dec of the 20M objects
#6
kmyi
opened
4 years ago
0
make algorithm that creates final prob_yufeng_{star,qso,gal}
#5
kmyi
closed
4 years ago
0
for catalogue in Milliquas, SDSS, GAIA:
#4
kmyi
closed
4 years ago
0
Go to HSC website and download HSC DR2 deep and ultra-deep catalogues (~20M objects) by SQL query.
#3
kmyi
closed
4 years ago
0
Read literature on seme-supervised classification.
#2
kmyi
closed
4 years ago
0
Prepare qusar labels. Have downloaded the dataset Milliquas which have label of QSO
#1
kmyi
closed
4 years ago
0