tpfister / caffe-heatmap

Caffe with heatmap regression & spatial fusion layers. Useful for any CNN image position regression task.

Home Page:http://www.robots.ox.ac.uk/~vgg/software/cnn_heatmap

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

How to crop the input image to get the same result as that in your paper?

xuxy09 opened this issue · comments

Hi, the test input images you provide are slightly different with the origin images in Pose-in-the-Wild dataset. Could you tell me how you did cropping? I can't find related information in your paper. Thanks!

Hi,

The labels and pre-cropped images for training and testing are now available.

http://tomas.pfister.fi/flic_train_cropped_multfact282.tgz
http://tomas.pfister.fi/flic_test_cropped_multfact282.tgz

@tpfister can you tell us please, how did you do the cropping of images for both test and train?
I need to know that because I will test your model on my own dataset.

commented

Hi, @xuxy09 @samouha
I am equally puzzled, have you been solved this?