hiroaki-santo/deep-photometric-stereo-network

This is the project page for our ICCVW 2017 paper 'Deep photometric stereo network' by Hiroaki Santo, Masaki Samejima, Yusuke Sugano, Boxin Shi, and Yasuyuki Matsushita.

deep-photometric-stereo-network's Issues

About rendered image

Hello, I found a mistake about the dataset.
Yesterday, I used the Dockerfile you provided to produce the dataset. But I found some materials image is very black and not normal, for example, the alumina-oxide in fig 1.

I opened this image in Matlab and I found its intensity is in [0,255]. Because the rendered image is uint16, I multiplied the rendered image by 255, it looks normal as follows：

However，It's not right for every material. The rendered image of blue-fabric is

and multiplied the rendered image by 255 is

it seems to overflow.
what is wrong?
Could you please provide the original dataset if it's convenient?
I am looking forward to your reply! Thank you very much

About mask and time

Hello，I have some questions about this paper。。。。

1. Each pixel of an image is treated as the training data in this paper. But I think there are lots of redundant data because of the existence of a mask. why ？
2. How long did it take to train this neural network?

Thank you！！

I want to use the estimated normal DiLiGenT results for the comparison

I want to use the estimated normal DiLiGenT results for the comparison, thank you.
I have email you (the email shown in paper) but haven't got the response...
Thankyou again!

About BRDFRead

Hello,
I am very interested in this project，but I got an error when compiling the project merl_brdf_database.I compiled on Mac os. There was no error in cmake, but the following error occurred during make.

Charlies-MacBook-Pro:merl_brdf_database charlie$ make
Scanning dependencies of target BRDFRead
[ 50%] Building CXX object CMakeFiles/BRDFRead.dir/BRDFRead.cpp.o
/Users/charlie/Documents/dpsn/deep-photometric-stereo-network/merl_brdf_database/BRDFRead.cpp:36:9: warning:
'M_PI' macro redefined [-Wmacro-redefined]
#define M_PI 3.1415926535897932384626433832795
^
/usr/include/math.h:703:9: note: previous definition is here
#define M_PI 3.14159265358979323846264338327950288 /* pi /
^
1 warning generated.
[100%] Linking CXX shared library BRDFRead.dylib
Undefined symbols for architecture x86_64:
"boost::numpy::initialize(bool)", referenced from:
init_module_BRDFRead() in BRDFRead.cpp.o
"boost::numpy::zeros(boost::python::tuple const&, boost::numpy::dtype const&)", referenced from:
read_brdfpy(char const) in BRDFRead.cpp.o
"boost::numpy::dtype boost::numpy::detail::get_float_dtype<64>()", referenced from:
read_brdfpy(char const*) in BRDFRead.cpp.o
"boost::python::detail::init_module(PyModuleDef&, void ()())", referenced from:
_PyInit_BRDFRead in BRDFRead.cpp.o
"boost::python::converter::object_manager_traitsboost::numpy::ndarray::get_pytype()", referenced from:
boost::python::detail::converter_target_type<boost::python::to_python_value<boost::numpy::ndarray const&> >::get_pytype() in BRDFRead.cpp.o
boost::python::detail::caller_arity<5u>::impl<boost::python::tuple ()(boost::numpy::ndarray, double, double, double, double), boost::python::default_call_policies, boost::mpl::vector6<boost::python::tuple, boost::numpy::ndarray, double, double, double, double> >::operator()(_object*, _object*) in BRDFRead.cpp.o
ld: symbol(s) not found for architecture x86_64
clang: error: linker command failed with exit code 1 (use -v to see invocation)
make[2]: *** [BRDFRead.dylib] Error 1
make[1]: *** [CMakeFiles/BRDFRead.dir/all] Error 2
make: *** [all] Error 2

I can't solve this problem. Do you have any suggestions?

About rendering_with_merl.py

Hello,
When I run "python rendering_with_merl.py",the error occured as follow.

Traceback (most recent call last):
File "rendering_with_merl.py", line 163, in
for n_path in n_png_paths[[1, 7]]: # This means blob02 and blob08
IndexError: index 1 is out of bounds for axis 1 with size 0

Can you solve it?Thank you.