[QUESTION] Invalid data type of cvcuda.nms() function

[ElinLiu0]

What is your question? Here is the question:the cvcuda.nms() function from documentation here need those below attributes to calculate:

• src boxes Tensor : nvcv.Tensor
• scores Tensor : nvcv.Tensor
• score Threshold : float
• iou Threshold : float I got all the input attributes with these construction code :

# Copy a clone for Tensor Construction
boxes_clone = boxes.copy()
scores_clone = scores.copy()
boxes_clone = torch.tensor(boxes_clone,device="cuda",dtype=torch.int32)
boxes_clone = nvcv.as_tensor(boxes_clone)
scores_clone = torch.tensor(scores_clone,device="cuda",dtype=torch.float32)
scores_clone = nvcv.as_tensor(scores_clone)


print(type(boxes_clone),type(scores_clone),type(self.confidenceThreshold),type(self.iouThreshold))
# Apply NMS CUDA Function
indices = cvcuda.nms(boxes_clone,scores)

print(indices)

But it raise me a InvokingError message like below:

nms(): incompatible function arguments. The following argument types are supported:
    1. (src: nvcv.Tensor, scores: nvcv.Tensor, score_threshold: float = 1.1920928955078125e-07, iou_threshold: float = 1.0, *, stream: Optional[nvcv.cuda.Stream] = None) -> nvcv.Tensor

Invoked with: <nvcv.Tensor shape=(164, 4) dtype=int32>, array([0.7586603 , 0.7097081 , 0.767958  , 0.78236204, 0.72838205,
       0.75135577, 0.75455606, 0.7190855 , 0.71297336, 0.82068956,
       0.7697949 , 0.78868264, 0.85104907, 0.7875968 , 0.7872595 ,
       0.86548656, 0.77008224, 0.7904344 , 0.7711759 , 0.7773221 ,
       0.76164293, 0.804789  , 0.8095949 , 0.7529364 , 0.8072941 ,
       0.8178994 , 0.703524  , 0.7738    , 0.7346774 , 0.7577837 ,
       0.7631812 , 0.8363519 , 0.76603943, 0.72979015, 0.7744066 ,
       0.77776486, 0.8415555 , 0.84321684, 0.8475255 , 0.74649525,
       0.7497457 , 0.7415061 , 0.8393519 , 0.8238603 , 0.82412285,
       0.74750257, 0.7030225 , 0.7291994 , 0.8311529 , 0.8128832 ,
       0.8061417 , 0.8527547 , 0.80295044, 0.79725283, 0.7942991 ,
       0.8023775 , 0.8073451 , 0.8075176 , 0.797567  , 0.7897532 ,
       0.7954175 , 0.810892  , 0.763571  , 0.8016494 , 0.7373098 ,
       0.77551764, 0.7829293 , 0.7970046 , 0.78473973, 0.79149765,
       0.75949925, 0.74192935, 0.8026274 , 0.7872242 , 0.7483539 ,
       0.8232488 , 0.79359096, 0.80303806, 0.8006276 , 0.735693  ,
       0.738103  , 0.81427705, 0.8062424 , 0.81192076, 0.8231721 ,
       0.8422905 , 0.7103702 , 0.74524415, 0.75137806, 0.8359127 ,
       0.83796084, 0.81999135, 0.7893438 , 0.8124736 , 0.76678663,
       0.7583429 , 0.70672303, 0.70196605, 0.716085  , 0.84689456,
       0.8341809 , 0.8348576 , 0.8163886 , 0.77195305, 0.7660649 ,
       0.876741  , 0.8607939 , 0.8236254 , 0.8343361 , 0.82574046,
       0.78387064, 0.78737396, 0.76560295, 0.87557435, 0.8643305 ,
       0.88600373, 0.8406993 , 0.8318555 , 0.86225355, 0.89372563,
       0.8801202 , 0.81545967, 0.8624695 , 0.8599481 , 0.8590247 ,
       0.88619626, 0.8880276 , 0.8758006 , 0.8741057 , 0.8697329 ,
       0.88414127, 0.8797039 , 0.8760199 , 0.88666004, 0.89366215,
       0.875077  , 0.75500464, 0.88037455, 0.8924499 , 0.88381225,
       0.8576473 , 0.86270386, 0.85533917, 0.8101083 , 0.8530666 ,
       0.8395133 , 0.84177613, 0.8450767 , 0.8457641 , 0.85781676,
       0.8484809 , 0.8510436 , 0.8571972 , 0.8495828 , 0.9065145 ,
       0.923493  , 0.91902345, 0.91615033, 0.9284622 , 0.9181302 ,
       0.92227095, 0.92974496, 0.91810983, 0.9094186 ], dtype=float32)
  File "/root/modelTraining/Apps/GPUClient/VideoProcessorGPU.py", line 102, in postprocess
    indices = cvcuda.nms(boxes_clone,scores)
  File "/root/modelTraining/Apps/GPUClient/test.py", line 14, in <module>
    processor.postprocess(
TypeError: nms(): incompatible function arguments. The following argument types are supported:
    1. (src: nvcv.Tensor, scores: nvcv.Tensor, score_threshold: float = 1.1920928955078125e-07, iou_threshold: float = 1.0, *, stream: Optional[nvcv.cuda.Stream] = None) -> nvcv.Tensor

Invoked with: <nvcv.Tensor shape=(164, 4) dtype=int32>, array([0.7586603 , 0.7097081 , 0.767958  , 0.78236204, 0.72838205,
       0.75135577, 0.75455606, 0.7190855 , 0.71297336, 0.82068956,
       0.7697949 , 0.78868264, 0.85104907, 0.7875968 , 0.7872595 ,
       0.86548656, 0.77008224, 0.7904344 , 0.7711759 , 0.7773221 ,
       0.76164293, 0.804789  , 0.8095949 , 0.7529364 , 0.8072941 ,
       0.8178994 , 0.703524  , 0.7738    , 0.7346774 , 0.7577837 ,
       0.7631812 , 0.8363519 , 0.76603943, 0.72979015, 0.7744066 ,
       0.77776486, 0.8415555 , 0.84321684, 0.8475255 , 0.74649525,
       0.7497457 , 0.7415061 , 0.8393519 , 0.8238603 , 0.82412285,
       0.74750257, 0.7030225 , 0.7291994 , 0.8311529 , 0.8128832 ,
       0.8061417 , 0.8527547 , 0.80295044, 0.79725283, 0.7942991 ,
       0.8023775 , 0.8073451 , 0.8075176 , 0.797567  , 0.7897532 ,
       0.7954175 , 0.810892  , 0.763571  , 0.8016494 , 0.7373098 ,
       0.77551764, 0.7829293 , 0.7970046 , 0.78473973, 0.79149765,
       0.75949925, 0.74192935, 0.8026274 , 0.7872242 , 0.7483539 ,
       0.8232488 , 0.79359096, 0.80303806, 0.8006276 , 0.735693  ,
       0.738103  , 0.81427705, 0.8062424 , 0.81192076, 0.8231721 ,
       0.8422905 , 0.7103702 , 0.74524415, 0.75137806, 0.8359127 ,
       0.83796084, 0.81999135, 0.7893438 , 0.8124736 , 0.76678663,
       0.7583429 , 0.70672303, 0.70196605, 0.716085  , 0.84689456,
       0.8341809 , 0.8348576 , 0.8163886 , 0.77195305, 0.7660649 ,
       0.876741  , 0.8607939 , 0.8236254 , 0.8343361 , 0.82574046,
       0.78387064, 0.78737396, 0.76560295, 0.87557435, 0.8643305 ,
       0.88600373, 0.8406993 , 0.8318555 , 0.86225355, 0.89372563,
       0.8801202 , 0.81545967, 0.8624695 , 0.8599481 , 0.8590247 ,
       0.88619626, 0.8880276 , 0.8758006 , 0.8741057 , 0.8697329 ,
       0.88414127, 0.8797039 , 0.8760199 , 0.88666004, 0.89366215,
       0.875077  , 0.75500464, 0.88037455, 0.8924499 , 0.88381225,
       0.8576473 , 0.86270386, 0.85533917, 0.8101083 , 0.8530666 ,
       0.8395133 , 0.84177613, 0.8450767 , 0.8457641 , 0.85781676,
       0.8484809 , 0.8510436 , 0.8571972 , 0.8495828 , 0.9065145 ,
       0.923493  , 0.91902345, 0.91615033, 0.9284622 , 0.9181302 ,
       0.92227095, 0.92974496, 0.91810983, 0.9094186 ], dtype=float32)

I don't know what's wrong with my input,any answer for that?

[AnimeDalit97]

I found some limitations in here

I think you need to convert the input tensor to int16

[ElinLiu0]

I found some limitations in here

I think you need to convert the input tensor to int16

thanks,but i still have some trouble need to deal with:because my environment will always raise a error:libstdc++-v3/src/c++11/codecvt.cc not found,even though i reinstall the WSL2 vm,this error still there,do you have any idea of that?

[bhaefnerNV]

Hi @ElinLiu0,

Thank you for your interest in CVCUDA.

It seems that you invoke the nms operator with scores, which, according to your example, is an array and not an nvcv.Tensor? I guess you could invoke nms with scores_clone, which is an nvcv.Tensor: cvcuda.nms(boxes_clone,scores) -> cvcuda.nms(boxes_clone,scores_clone)

[bhaefnerNV]

Hi @ElinLiu0, we will close this issue due to inactivity. Please feel free to reopen, if the problem persists.

[ElinLiu0]

Hi @ElinLiu0, we will close this issue due to inactivity. Please feel free to reopen, if the problem persists.

Thanks buddy,i've already use CVCUDA accelerate my YoloV8 process pipeline: https://github.com/ElinLiu0/StudentBehaviorDetection/blob/gpuPipe/VideoProcessorGPUCVCUDA.py