Dear @shurans ,

After following the guideline and running the Matlab command 'dss_marvin_script(0,100,1,[] ,1,'RPN_NYU',1,[],0,0);', I meet an error:

Error using load
Unable to read file '/n/fs/modelnet/deepDetect/code/detector_3d/hasorientation': no such file or
directory.
Error in dss_preparelist (line 8)
load('/n/fs/modelnet/deepDetect/code/detector_3d/hasorientation');
Error in dss_marvin_script (line 155)
    [boxfile_traincnn] =
    dss_preparelist('train',NYUonly,opt.proposal_dir,opt.proposal_dir,proposal_only,cls,opt.box_reg,opt.axisAlign,opt.fullbox,opt.BoxperImage,0,0); 

However I don't know where to download this folder. Could you please give some hint~

Thanks a lot
Mengqi

Hi,what's the structure of the dataset SUNRGBDMeta.mat,i open it by matlab,but can't understand it clearly,if i do a .mat file by myself,what can i do?thanks !

Hi @shurans ,

It seems that downloading precomputed regoin proposal dataset by running script: ('../proposal','http://dss.cs.princeton.edu/Release/proposal/','.tensor')failed. Could you check if that's the correct URL? Thank you.

I use apt-get to install libsomesa6-dev on ubuntu 15.10 with GTX 980ti, but I found the system is not work. So there is any requirement about Mesa3D? Thank you very much

Hi @shurans ,

It seems that downloading precomputed regoin proposal for NYU and SUNRGBD dataset by running script: downloadData('../image','http://dss.cs.princeton.edu/Release/sunrgbd_dss_data/','.tensor') failed becasue of Error 404. Could you check if that's the correct URL? Thank you.

Hi @shurans,

I am having an issue when running following line in dss_prepareAnchorbox load('./external/SUNRGBDtoolbox/Metadata/SUNRGBDMeta_tight_Yaw.mat');

I also looked inside following url that this file doesn't exist.

http://dss.cs.princeton.edu/Release/

Could you please help me where I can download this file? Thanks.

when I run this script:
downloadData('../sunrgbd_dss_data','http://dss.cs.princeton.edu/Release/sunrgbd_dss_data/','.bin');
there is an error :
bash: 未预期的符号 `'../sunrgbd_dss_data','http://dss.cs.princeton.edu/Release/sunrgbd_dss_data/','.bin'' 附近有语法错误.
how should I do?

Hi,

Are you computing Directional TSDF anywhere in this package. I am unable to find it. Please let me know how do you generate tsdfSpace.ply

HI shurans, Once i have done with training the proposal model,then i ran the RPN_Extract() function to get the proposals. I found there no source of
cls_score_.tensor,
cls_score_1_.tensor,
box_pred_reshape_.tensor,
box_pred_reshape_1_.tensor
and so on .
There are so many missing files to get the proposal result.

Hi shurans,
I found there was another paper about sliding shape, sliding shape(2014ECCV), there is a section about feature visualization. Can you give me some detail about your 4 features(point density,3D normal, TSDF, 3D shape),what is the physical or geometric meaning?
Thank u very much.

Hello again @shurans,

As far as I understand when we run dss_prepareAnchorbox(), we get the dataset which is given in (http://dss.cs.princeton.edu/Release/RPNdata_mulit/SUNRGBD/kv1/NYUdata) am I right? Because of different path issues I could not run dss_prepareAnchorbox(). Therefore, if it just produces the data in the above link, I do not need to run dss_prepareAnchorbox() and just can use the already prepared dataset at [http://dss.cs.princeton.edu/Release/RPNdata_mulit/SUNRGBD/kv1/NYUdata] .Am I right?

Thanks,

Hi @shurans,

In DSSopt_RPN.json, you use RPN_NYU_train.list and RPN_NYU_test.list. However, it does not exist in http://dss.cs.princeton.edu/Release folder. Where can we find RPN_NYU_train.list and RPN_NYU_test.list files?

Thx

Does this framework support multi-gpu?

I wonder the size of the your dataset. what is the size of whole dataset? I start the downloading process in the morning, it is still downloading.

Hi Shuran,

I found there are some configuration needed in RPN_Extract function to point to the correct path. There are:

/home/shurans/deepDetectLocal/RPNresult/
/n/fs/modelnet/deepDetect/RPNresult/
/n/fs/modelnet/deepDetect/proposal/RPN_
/n/fs/modelnet/SUN3DV2/prepareGT/cls.mat

Where can I set these folder correctly and point to the data I downloaded through the instruction? I have run dss_prepareAnchorBox already. Thank you.

Hi,
I see that there are a lot of hard coded paths in your code.
I think , the initial hard coded path is in dss_initPath.m file.
Please remove the hard coded paths from the code.

Also please , let us know how can i run the model to test on the testing data.

Thanks

Hi @shurans

How can I read the .tensor file in HHA directory?

I change the number-anchor,sizegroup and resharp,and however the resharp in marvin.hpp told me uncorrect and cuda error 11.

Hi Shurans,
my environment:
cuda7.5
cudnn-7.0-linux-x64-v3.0-rc
when I run ./linux.sh, there is an error as follow:

stc@stc:/Downloads/marvin$ ./linux.sh
rm: cannot remove ‘marvin’: No such file or directory
marvin.hpp(3662): error: identifier "cudnnAddTensor_v3" is undefined

1 error detected in the compilation of "/tmp/tmpxft_0000286d_00000000-9_marvin.cpp1.ii".

Is there some problem with my environment? Could you update the version based on cuda7.5 and cudnn5.1.

Hi, @shurans ,

I follow the setup guide, but came accross the following compilation error, any pre-requisite packages to install beforehand?

root@milton-All-Series:~/DeepSlidingShape/marvin# ./linux.sh
In file included from dss.hpp:1:0,
from marvin.hpp:5155,
from marvin.cu:1:
render_3d.hpp:7:23: fatal error: GL/osmesa.h: No such file or directory
#include <GL/osmesa.h>
^
compilation terminated.

THX~

I cant run linux.sh file, there exist some problems. Could you check it ?
_input_

../DeepSlidingShape/marvin(master)$ ./linux.sh 

_error_

rm: cannot remove 'marvin': No such file or directory
/usr/lib/gcc/x86_64-linux-gnu/5/include/mwaitxintrin.h(36): error: identifier "__builtin_ia32_monitorx" is undefined

/usr/lib/gcc/x86_64-linux-gnu/5/include/mwaitxintrin.h(42): error: identifier "__builtin_ia32_mwaitx" is undefined

/usr/include/c++/5/functional(78): error: class "std::thread" has no member "result_type"
          detected during:
            instantiation of class "std::_Maybe_get_result_type<_Functor, void> [with _Functor=std::thread]" 
(86): here
            instantiation of class "std::_Weak_result_type_impl<_Functor> [with _Functor=std::thread]" 
(184): here
            instantiation of class "std::_Weak_result_type<_Functor> [with _Functor=std::thread]" 
(264): here
            instantiation of class "std::_Reference_wrapper_base_impl<true, true, _Tp> [with _Tp=std::thread]" 
(283): here
            instantiation of class "std::_Reference_wrapper_base<_Tp> [with _Tp=std::thread]" 
(399): here
            instantiation of class "std::reference_wrapper<_Tp> [with _Tp=std::thread]" 
/usr/include/c++/5/future(1638): here

/usr/include/c++/5/functional(266): error: class "std::thread" has no member "argument_type"
          detected during:
            instantiation of class "std::_Reference_wrapper_base_impl<true, true, _Tp> [with _Tp=std::thread]" 
(283): here
            instantiation of class "std::_Reference_wrapper_base<_Tp> [with _Tp=std::thread]" 
(399): here
            instantiation of class "std::reference_wrapper<_Tp> [with _Tp=std::thread]" 
/usr/include/c++/5/future(1638): here

/usr/include/c++/5/functional(267): error: class "std::thread" has no member "first_argument_type"
          detected during:
            instantiation of class "std::_Reference_wrapper_base_impl<true, true, _Tp> [with _Tp=std::thread]" 
(283): here
            instantiation of class "std::_Reference_wrapper_base<_Tp> [with _Tp=std::thread]" 
(399): here
            instantiation of class "std::reference_wrapper<_Tp> [with _Tp=std::thread]" 
/usr/include/c++/5/future(1638): here

/usr/include/c++/5/functional(268): error: class "std::thread" has no member "second_argument_type"
          detected during:
            instantiation of class "std::_Reference_wrapper_base_impl<true, true, _Tp> [with _Tp=std::thread]" 
(283): here
            instantiation of class "std::_Reference_wrapper_base<_Tp> [with _Tp=std::thread]" 
(399): here
            instantiation of class "std::reference_wrapper<_Tp> [with _Tp=std::thread]" 
/usr/include/c++/5/future(1638): here

marvin.hpp(3381): error: too few arguments in function call

marvin.hpp(3705): error: argument of type "cudnnAddMode_t" is incompatible with parameter of type "const void *"

marvin.hpp(3706): error: argument of type "const void *" is incompatible with parameter of type "cudnnTensorDescriptor_t"

marvin.hpp(3709): error: argument of type "const void *" is incompatible with parameter of type "cudnnTensorDescriptor_t"

marvin.hpp(3711): error: too many arguments in function call

marvin.hpp(3728): error: argument of type "const void *" is incompatible with parameter of type "cudnnConvolutionBwdDataAlgo_t"

marvin.hpp(3729): error: argument of type "float *" is incompatible with parameter of type "size_t"

marvin.hpp(3729): error: too few arguments in function call

marvin.hpp(3756): error: argument of type "float *" is incompatible with parameter of type "cudnnConvolutionBwdFilterAlgo_t"

marvin.hpp(3757): error: argument of type "float *" is incompatible with parameter of type "size_t"

marvin.hpp(3757): error: too few arguments in function call

/usr/include/c++/5/bits/stl_iterator_base_types.h(154): error: name followed by "::" must be a class or namespace name
          detected during:
            instantiation of class "std::__iterator_traits<_Iterator, void> [with _Iterator=int]" 
(163): here
            instantiation of class "std::iterator_traits<_Iterator> [with _Iterator=int]" 
dss.hpp(1321): here

/usr/include/c++/5/bits/stl_iterator_base_types.h(155): error: name followed by "::" must be a class or namespace name
          detected during:
            instantiation of class "std::__iterator_traits<_Iterator, void> [with _Iterator=int]" 
(163): here
            instantiation of class "std::iterator_traits<_Iterator> [with _Iterator=int]" 
dss.hpp(1321): here

/usr/include/c++/5/bits/stl_iterator_base_types.h(156): error: name followed by "::" must be a class or namespace name
          detected during:
            instantiation of class "std::__iterator_traits<_Iterator, void> [with _Iterator=int]" 
(163): here
            instantiation of class "std::iterator_traits<_Iterator> [with _Iterator=int]" 
dss.hpp(1321): here

/usr/include/c++/5/bits/stl_iterator_base_types.h(157): error: name followed by "::" must be a class or namespace name
          detected during:
            instantiation of class "std::__iterator_traits<_Iterator, void> [with _Iterator=int]" 
(163): here
            instantiation of class "std::iterator_traits<_Iterator> [with _Iterator=int]" 
dss.hpp(1321): here

/usr/include/c++/5/bits/stl_iterator_base_types.h(158): error: name followed by "::" must be a class or namespace name
          detected during:
            instantiation of class "std::__iterator_traits<_Iterator, void> [with _Iterator=int]" 
(163): here
            instantiation of class "std::iterator_traits<_Iterator> [with _Iterator=int]" 
dss.hpp(1321): here

22 errors detected in the compilation of "/tmp/tmpxft_00001620_00000000-9_marvin.cpp1.ii".

Hi @shurans ,
When I loaded SUNRGBDMetadata.mat in dss_prepareAnchor(),I found that it recorded the paths of depth and rgb image. But I can't find any png data in the directory. Do we need to download the data from the original SUNDRGBD website? or the .bin file in "sunrgbd_dss_data/SUNRGBD/kv1/NYUdata/*.bin" are datasets?
If yes, what is the information in .bin file? How can we convert them to depth and rgb images?

Thanks a lot!

I am not an expert on 3D scene understanding. Sorry if my question is too low ...
But I have a question on the data used for this work. My understanding on RGB-D images is they are composed of two parts, one is the normal 2D RGB image and the other is the depth image which is also 2D actually. How to get the 3D image from them and use the 3D image as the input of the neural network? And if there are occlusions in 2D images, are there occlusions in the obtained 3D images?

I am currently working on a research project that intends to use Deep Sliding Shapes as a baseline comparison. However I use Keras, as that is more widely supported in comparison with marvin. Would you consider converting the model definition and the pretrained weights to Keras? It would make the usage of Deep Sliding Shapes so much easier.

I can't find the bb2dOverlap function or file in the code, but it is used several times. Thanks. @shurans

Hi, it seems that the file_list of the 3D scenes is missing.. could you please provide an example of file_list (such as RPN_NYU.list, defined in RPNDataLayer of DSSopt_RPN_extractfea_po.json)? Thanks.