ttgeng233 / UnAV

Dense-Localizing Audio-Visual Events in Untrimmed Videos: A Large-Scale Benchmark and Baseline (CVPR 2023)
https://unav100.github.io
MIT License
52 stars 3 forks source link

missing nms_cpu.cpp file #1

Closed linxi1158 closed 1 year ago

linxi1158 commented 1 year ago

I greatly appreciate the work you have done. However, during the setup of this project, I encountered two issues:

  1. Environment configuration bug: according to the content in utils/setup.py, there is a missing file './csrc/nms_cpu.cpp' in this project.
setup(
    name='nms_1d_cpu',
    ext_modules=[
        CppExtension(
            name = 'nms_1d_cpu',
            sources = ['./csrc/nms_cpu.cpp'],
            extra_compile_args=['-fopenmp']
        )
    ],
    cmdclass={
        'build_ext': BuildExtension
    }
)
  1. Parameter passing bug: In multimodal_meta_archs.py, on line 266, the DependencyBlock class is called without the pyramid_level parameter being passed.
if self.use_dependency:
            self.dependency_block = make_dependency_block(
                'DependencyBlock',
                **{
                    'in_channel' : embd_dim*2,
                    'n_embd' : 128,  
                    'n_embd_ks' : embd_kernel_size,
                    'num_classes' : self.num_classes,
                    'pyramid_level' : backbone_arch[-1] + 1, # n_head?
                    'path_pdrop' : self.train_droppath,
                }
            )
linxi1158 commented 1 year ago

bug 1 : I use https://github.com/TuanTNG/TemporalMaxer/tree/master/libs/utils/csrc to solve ! but bug 2?

ttgeng233 commented 1 year ago

Sorry for making confusion! I have added the file './csrc/nms_cpu.cpp' and delete 'pyramid_level' on line 266. Thank you for your reminder! If you have other questions, please feel free to tell me.

linxi1158 commented 1 year ago

OK