mnc doesnt recognize any object

[limorbagizada]

I try to run the demo, with the suggested trained MNC model and the original 5 images, but the output images getting from the algorithm are black without any recognized objects in the picture.I notice that the im_detect function return very low scores, lower than 0.1
I didn't change anything in the code, what can be the problem?

[Guominyingxiongququ]

I have the same issue...

[PangolinLE314]

I have the same issue when working with GTX1080 GPU and cuDNN v4. Build with CuDNN V5 solved this problem.

[AndreiBarsan]

Alternatively, a dirty solution is to disable cuDNN. In my case, inference still runs reasonably fast wihthout cuDNN.

@PangolinLE314 I sadly could not build MNC's fork of Caffe with cuDNN 5. Did you have to do anything in particular to get it to build with cuDNN 5?

[PangolinLE314]

@AndreiBarsan
Here is the solution from manipopopo:

cd caffe-fast-rcnn  
git remote add caffe https://github.com/BVLC/caffe.git  
git fetch caffe  
git merge -X theirs caffe/master  

Remove self_.attr("phase") = static_cast(this->phase_); from include/caffe/layers/python_layer.hpp after merging.

rbgirshick/py-faster-rcnn#237 (comment)

[AndreiBarsan]

@PangolinLE314 awesome! That worked! I could build with cuDNN 5, all the Caffe tests passed, and inference is now considerably faster and uses less memory. Thank you very much!

[deartonym]

@PangolinLE314
I have exactly the same problem and thanks PangolinLE.

[mldm4]

Hi, I am having this problem working with cuDNN4. I have done what @PangolinLE314 proposes, but then I got pycaffe could not import NCCL nor Timer. If anyone has a clue about what I could do I would be extremely grateful. Cheers!

[mtbui2010]

Hi, I have the same problem with @mldm4. Please give me a guide if you did solve this issue. I am using cuda 6.5, cudnn 4.0 on the ubuntu 14.04.