Hi, I study physics in Germany and i really enjoy your videos! Honestly spoken, i doubt that the network can generalize Denoising to other pictures, even if they are similar to the used one s. I think that the memory of the decoder, there are 50*100+100*128*128 weights (+biases), is more than enough to store the 10 pictures. I will try to put in some other pictures Greedings
Hi, Professor Jeff! :) I’m just wondering-after we obtained the most important features from the bottleneck of our trained neural network, is it possible to apply the denoising capability of the autoencoder to a live feed video that is somewhat highly correlated to the training images? Will this be better, or even recommended, instead of using traditional denoising filters of OpenCV for real-time videos? I’d love to learn more from your expertise and advices as I explore this topic further. Thank you for the insightful explanation and demo by the way! Subscribed! :)
This is not something that I've tried, but it sounds like a valid approach. I've added this idea to my future video list, I want to do more "video" videos.
Professor Heaton, I am trying the single image auto-encoder and happens to find out the accuracy is always 0 while the loss decreased from 12481.3857 to near 0.(after 200 epochs) Did I set the model wrong?( I used the same set up and Sequential model like yours in the code) Thank you! Great video!
model.compile(optimizer='adam', loss='mean_squared_error', metrics=['accuracy']) history = model.fit(img_array,img_array,verbose=1,epochs=200) This is how I set up the training
hello .can we use autoencoders to generate a more accurate dataset where the nan values are gone.in other words use autoencoders to fill the missing values thanks
Thank you so much for these videos and all that hard work Jeff, this is really opening up a whole new world of coding to me, really appreciate it! :)
Hi, I study physics in Germany and i really enjoy your videos! Honestly spoken, i doubt that the network can generalize Denoising to other pictures, even if they are similar to the used one
s. I think that the memory of the decoder, there are 50*100+100*128*128 weights (+biases), is more than enough to store the 10 pictures. I will try to put in some other pictures
Greedings
Hi, Professor Jeff! :)
I’m just wondering-after we obtained the most important features from the bottleneck of our trained neural network, is it possible to apply the denoising capability of the autoencoder to a live feed video that is somewhat highly correlated to the training images?
Will this be better, or even recommended, instead of using traditional denoising filters of OpenCV for real-time videos?
I’d love to learn more from your expertise and advices as I explore this topic further. Thank you for the insightful explanation and demo by the way! Subscribed! :)
This is not something that I've tried, but it sounds like a valid approach. I've added this idea to my future video list, I want to do more "video" videos.
Professor Heaton, I am trying the single image auto-encoder and happens to find out the accuracy is always 0 while the loss decreased from 12481.3857 to near 0.(after 200 epochs) Did I set the model wrong?( I used the same set up and Sequential model like yours in the code) Thank you! Great video!
model.compile(optimizer='adam', loss='mean_squared_error', metrics=['accuracy'])
history = model.fit(img_array,img_array,verbose=1,epochs=200)
This is how I set up the training
hello .can we use autoencoders to generate a more accurate dataset where the nan values are gone.in other words use autoencoders to fill the missing values
thanks
Jeff, you are next level. buddy
Thanks! Appreciate it.