Resolution of output disparity map should be 512x512 in case of synthetic LF data

chshin10 / epinet

EPINET: A Fully-Convolutional Neural Network using Epipolar Geometry for Depth from Light Field Images

MIT License

88 stars 39 forks source link

Resolution of output disparity map should be 512x512 in case of synthetic LF data #5

Open jackpt8671 opened 6 years ago

jackpt8671 commented 6 years ago

Hi Changha,

Very thankful for your github contribution code. Can you confirm the resolution of disparity map ? In your code, it generates 490x490 one but 512x512 is more correct, isnt it? I ask this question because the ground truth of synthetic LF has resolution 512x512. I want to make a fair comparison.

Have a good day, btw.

chshin10 commented 6 years ago

Yes, our output size is 490x490 pixels. Since we didn't use padding in convolution layers, our output size was reduced. In HCI benchmark site, they also compare the output result in (482, 482) pixels. I think you can use their source code for a fair comparison. (https://github.com/lightfield-analysis/evaluation-toolkit#1-evaluate-light-field-algorithms)