Posts

Post is pinned.Post has attachment

Public

Deep Learning in Computer Vision @ Deep Learning in Action Meetup !

Post has attachment

Hi Dr. Dansereau,

I would like to extract the intrinsic camera parameters from the H matrix. However in the paper, H matrix is derived by multiplying 6 3x3 matrices and the final H matrix is 5x5. First I would like to know how this is possible and I also want to know the exact values for dmu, dM and focal length of lenses. Is this possible to retrieve these values from the H matrix?

Thanks,

Caner

I would like to extract the intrinsic camera parameters from the H matrix. However in the paper, H matrix is derived by multiplying 6 3x3 matrices and the final H matrix is 5x5. First I would like to know how this is possible and I also want to know the exact values for dmu, dM and focal length of lenses. Is this possible to retrieve these values from the H matrix?

Thanks,

Caner

Hi everyone,

Has anybody worked with the warped depth maps, produced with Lytro Power Tool Beta? I would like to know how to process them to have a nice visualization. Output size is 571x376 whereas the original image has 2022x1404 resolution.

What is the bit depth of uint6 depth image files?

Thanks!

Has anybody worked with the warped depth maps, produced with Lytro Power Tool Beta? I would like to know how to process them to have a nice visualization. Output size is 571x376 whereas the original image has 2022x1404 resolution.

What is the bit depth of uint6 depth image files?

Thanks!

Hi Dr. Danseareau,

Thank you for this awesome toolbox! I am processing light fields to refocus the images with shift-sum filter. I would be glad if you could lead me to any source to understand better about the relation between the slope and the focal length. In your LFDemoBasicFiltLytroF01 script you are setting the slope to x/9. Is this because you have a 9x9 lenslet size?

Looking forward to your response!

Thanks!

Caner

Thank you for this awesome toolbox! I am processing light fields to refocus the images with shift-sum filter. I would be glad if you could lead me to any source to understand better about the relation between the slope and the focal length. In your LFDemoBasicFiltLytroF01 script you are setting the slope to x/9. Is this because you have a 9x9 lenslet size?

Looking forward to your response!

Thanks!

Caner

Post has shared content

This is super cool work - generating RGB images from normal maps and more.

http://arxiv.org/pdf/1603.05631v1.pdf

http://arxiv.org/pdf/1603.05631v1.pdf

Post has shared content

We propose a simple technique to expose the implicit attention of CNNs on the image. It highlights the most informative image regions relevant to the predicted class. You could get attention-based model instantly by tweaking your own CNN a little bit. Take a look at this video to see how CNN shifts its attention over time The work is published at CVPR'16. All models and code are available at http://cnnlocalization.csail.mit.edu/

Post has attachment

Impressive! Colorful Image Colorization with deep learning: http://arxiv.org/pdf/1603.08511.pdf

Post has attachment

CAFFE Debugger...

http://www.nvidia.com/object/nsight.html

http://www.nvidia.com/object/nsight.html

Post has attachment

Why Xavier in weight initialization: http://andyljones.tumblr.com/post/110998971763/an-explanation-of-xavier-initialization

#deeplearning #neuralnetworks #weightinitialization #caffe #Xavier

#deeplearning #neuralnetworks #weightinitialization #caffe #Xavier

Wait while more posts are being loaded