test failed

[mawenwei]

Duringtesting two models, I encountered a question that results in failure. The question is: F0831 14:32:22.325904 15088 layer_factory.hpp:81] Check failed : registry.count(type) == 1(0 vs. 1) Unknown layer type: BilinearInterpolation(known type: AbsVal, Accuracy, AryMax, BNLL, ....., Tile, WindowData)

*** Check failure stack trace:*** Aborted

[mawenwei]

I don't have nvidia-docker, I have installed successfully all software which you listed

[ctensmeyer]

You need to use my custom fork of caffe: https://github.com/ctensmeyer/caffe

This has the implementation for the BilinearInterpolation Layer.

[Mypathissional]

Hi @ctensmeyer , Is there a possibility to use other layer or maybe combination of other layers in Caffe to replace BilinearInterpolation Layer (maybe deconv or by bilinear filter) ?

I will appreciate any help.

[ctensmeyer]

You could use a deconv layer to perform bilinear upsampling with the proper weights. That should work for input images whose dimensions are multiples of 16. That way, when you downsample and then upsample by a factor of 16, you recover the original input size for the output.

On Thu, Sep 27, 2018 at 10:18 AM Mypathissional [email protected] wrote:

Hi @ctensmeyer https://github.com/ctensmeyer , Is there a possibility to use other layer or maybe combination of other layers in Caffe to replace BilinearInterpolation Layer (maybe deconv) ?

I will appreciate any help.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/ctensmeyer/binarization_2017/issues/2#issuecomment-425154490, or mute the thread https://github.com/notifications/unsubscribe-auth/AC2mpK5CODvdtQbf2A2TPKBmsc2HUWJdks5ufPprgaJpZM4WUrwH .