SLIDE 1
Noise2Self: Blind Denoising by Self-Supervision
Joshua Batson
Noise2Self: Blind Denoising by Self-Supervision Joshua Batson Loc - - PowerPoint PPT Presentation
Noise2Self: Blind Denoising by Self-Supervision Joshua Batson Loc - - PowerPoint PPT Presentation
Noise2Self: Blind Denoising by Self-Supervision Joshua Batson Loc Royer Noisy Data Supervision Supervision Supervision Self-Supervision? Self-Supervision? Self-Supervision? Self-Supervision? Self-Supervision? Self-Supervision?
SLIDE 2
SLIDE 3
Supervision
SLIDE 4
Supervision
SLIDE 5
Supervision
SLIDE 6
Self-Supervision?
SLIDE 7
Self-Supervision?
SLIDE 8
Self-Supervision?
SLIDE 9
Self-Supervision?
SLIDE 10
Self-Supervision?
SLIDE 11
Self-Supervision?
SLIDE 12
Self-Supervision?
SLIDE 13
Self-Supervision?
SLIDE 14
Self-Supervision?
SLIDE 15
Single-Image Self-Supervised CNN Training
SLIDE 16
Single-Image Self-Supervised CNN Training
SLIDE 17
Single-Image Self-Supervised CNN Training
SLIDE 18
J-invariant Deep CNN
SLIDE 19
J-invariant Deep CNN
SLIDE 20
Plus...
Code
for i, batch in enumerate(data_loader): noisy_images = batch input, mask = masker.mask(noisy_images, i)
- utput = model(input)
loss = loss_function(output*mask, noisy_images*mask)
Definitions Gaussian Processes Single-Cell Sequencing Theorems
4 Mpo Gene erythroid cells myeloid cells stem cells
- ptimal
Matrix Factorization
github.com/czbiohub/noise2self poster #118
SLIDE 21
donut
noisy noisy
SLIDE 22
donut
r=1 r=2 r=3 r=4 r=5
noisy noisy
SLIDE 23
r=5
MSE Radius of median filter self-supervised ground truth
donut
r=1 r=2 r=3 r=4 r=5
noisy noisy