You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
there are great examples in caffe website such as RCNN, but it seems regression is missing. I want to train a network which as output gives me the optical flow map of an image using Euclidean Loss layer and a specific accuracy layer for regression. also this network has two loss functions as multitask loss. one for flow magnitude and another one for optical flow vector which the output for latter case is a two channel matrix.
Right Now I have some images and their corresponding flow maps as labels. I guess it would be great if in the next version of caffe such an example would be added.
Currently I have 500 training images as start with dimension (500,500,3) with values 0 to 255. and the labels of flow magnitudes are with dimension (500,500). I put have used euclidean loss and fully convolutional layers as well as deconvolutional layer. I hope everything has been done in a good way. so this example can be put in caffe website as an regression example in optical flow predicition.
The text was updated successfully, but these errors were encountered:
Hi,
there are great examples in caffe website such as RCNN, but it seems regression is missing. I want to train a network which as output gives me the optical flow map of an image using Euclidean Loss layer and a specific accuracy layer for regression. also this network has two loss functions as multitask loss. one for flow magnitude and another one for optical flow vector which the output for latter case is a two channel matrix.
Right Now I have some images and their corresponding flow maps as labels. I guess it would be great if in the next version of caffe such an example would be added.
Currently I have 500 training images as start with dimension (500,500,3) with values 0 to 255. and the labels of flow magnitudes are with dimension (500,500). I put have used euclidean loss and fully convolutional layers as well as deconvolutional layer. I hope everything has been done in a good way. so this example can be put in caffe website as an regression example in optical flow predicition.
The text was updated successfully, but these errors were encountered: