Actually, I don't mind when you type the code in. It slows the process down and allows me to think about what you are actually doing. Have you set up a github account yet?
Github account has been created and I started to slowly upload code, after confirming it works on my system one last time. Here is the link: github.com/bnsreenu/python_for_microscopists
Thank you very much sir this is the only channel i found very usefull for deep learning sir can you pls suggest the latest de noising methods in deep learning.
Hi, thanks a lot for this. I am trying to denoise some medical ct data using this technique. I used random noise to make the noisy data from clean data. Although I am getting ultra low loss values (e-7), my accuracy is 0.27! Any idea why is that and how I can improve it? I have 600 pictures as train data and 400 as test data. For the random noise I used the standard deviation of only 0.02.
Congrats on the amazing content But I'm wondering this. In this example, firstly, a noisy image is obtained by adding noise to the existing image, and then the encoder is trained to eliminate the added noise from the noisy image. Therefore, the encoder is trained to eliminate random noise from any image. Alright Can a structure that can only denoise random noise be effective in real applications? In other words, if we want to eliminate noise from the data set we want to use, how can we eliminate other noises other than random noise by using an encoder.
Many denoising algorithms assume the noise to be random, including deep learning approaches such as Noise2Void. The white noise you get from many sources is indeed random. If you have structured noise, you can try cleaning it up by converting your image into Fourier space. In other words, you can try frequency domain filters rather than spatial denoising filters.
i am preparing for my placements and damn i found your channel its pure gold ! Thank ypu so much
very very nice material thanks sir
Thank you ! This channel is really amazing.
Glad you enjoy it!
How are you implementing the latent space without reparameterization trick?
Thank you! This video was very helpful.
Actually, I don't mind when you type the code in. It slows the process down and allows me to think about what you are actually doing.
Have you set up a github account yet?
Github account has been created and I started to slowly upload code, after confirming it works on my system one last time. Here is the link: github.com/bnsreenu/python_for_microscopists
Thank you very much sir this is the only channel i found very usefull for deep learning sir can you pls suggest the latest de noising methods in deep learning.
May be this video helps... ua-cam.com/video/yO15IISXA1Y/v-deo.html
Hi, thanks a lot for this. I am trying to denoise some medical ct data using this technique. I used random noise to make the noisy data from clean data. Although I am getting ultra low loss values (e-7), my accuracy is 0.27! Any idea why is that and how I can improve it? I have 600 pictures as train data and 400 as test data. For the random noise I used the standard deviation of only 0.02.
increase train data size.
Congrats on the amazing content But I'm wondering this. In this example, firstly, a noisy image is obtained by adding noise to the existing image, and then the encoder is trained to eliminate the added noise from the noisy image. Therefore, the encoder is trained to eliminate random noise from any image. Alright Can a structure that can only denoise random noise be effective in real applications? In other words, if we want to eliminate noise from the data set we want to use, how can we eliminate other noises other than random noise by using an encoder.
Many denoising algorithms assume the noise to be random, including deep learning approaches such as Noise2Void. The white noise you get from many sources is indeed random. If you have structured noise, you can try cleaning it up by converting your image into Fourier space. In other words, you can try frequency domain filters rather than spatial denoising filters.
@@DigitalSreeni Your ideas are very valuable. Good luck.take care