opencv - Assertion failed in ORB CUDA: Error with GpuMat and ROI in detectAndComputeAsync - Stack Overflow

admin管理员组

文章数量:1026062

I am trying to implement ORB with CUDA in OpenCV. I created an ORB object and am attempting to run orb.detectAndComputeAsync() on a GpuMat. However, I am encountering the following error:

cv2.error: OpenCV(4.10.0-dev) ... error: (-215:Assertion failed) 
   0 <= roi.x && 0 <= roi.width && roi.x + roi.width <= m.cols &&
   0 <= roi.y && 0 <= roi.height && roi.y + roi.height <= m.rows
   in function 'cv::cuda::GpuMat::GpuMat'

I am trying to upload a grayscale image to the GPU and use CUDA's ORB for keypoint detection and descriptor computation. Below is the Python code that initializes ORB and attempts to run the detectAndComputeAsync() function:

# Initialize ORB for CUDA
orb = cv2.cuda.ORB.create(
    nfeatures=1500, 
    scaleFactor=1.2,
    nlevels=8,
    edgeThreshold=31,
    firstLevel=0,
    WTA_K=4,
    scoreType=cv2.ORB_HARRIS_SCORE
)

print(f"GPU Image Channels: {gpu_image.channels()}, Type: {gpu_image.type()}")

# Detect and compute descriptors
keypoints, gpu_descriptors = orb.detectAndComputeAsync(gpu_image, None)

The gpu_image I am passing has a size of (180, 290), a channel of 1, and a type of 0

Relevant C++ code referred by the error: The error message points to an assertion failure in the following C++ code within OpenCV's CUDA implementation:

cv::cuda::GpuMat::GpuMat(const GpuMat& m, Rect roi) :
    flags(m.flags), rows(roi.height), cols(roi.width),
    step(m.step), data(m.data + roi.y * step), refcount(m.refcount),
    datastart(m.datastart), dataend(m.dataend),
    allocator(m.allocator)
{
    data += roi.x * elemSize();

    CV_Assert(0 <= roi.x && 0 <= roi.width &&
              roi.x + roi.width <= m.cols &&
              0 <= roi.y && 0 <= roi.height &&
              roi.y + roi.height <= m.rows);

    if (refcount)
        CV_XADD(refcount, 1);

    if (rows <= 0 || cols <= 0)
        rows = cols = 0;
    updateContinuityFlag();
}

Additional Details: Environment: OpenCV 4.10.0-dev with CUDA support. GPU Image Size: (180, 290) (Height x Width). Image Type: Single-channel (grayscale) with type 0

What I've tried: I modified the code to ensure that the ROI is within the bounds of the GPU image, but I still get the error.

Any guidance on what could be causing this issue or how to fix it would be greatly appreciated.

I am trying to implement ORB with CUDA in OpenCV. I created an ORB object and am attempting to run orb.detectAndComputeAsync() on a GpuMat. However, I am encountering the following error:

cv2.error: OpenCV(4.10.0-dev) ... error: (-215:Assertion failed) 
   0 <= roi.x && 0 <= roi.width && roi.x + roi.width <= m.cols &&
   0 <= roi.y && 0 <= roi.height && roi.y + roi.height <= m.rows
   in function 'cv::cuda::GpuMat::GpuMat'

I am trying to upload a grayscale image to the GPU and use CUDA's ORB for keypoint detection and descriptor computation. Below is the Python code that initializes ORB and attempts to run the detectAndComputeAsync() function:

# Initialize ORB for CUDA
orb = cv2.cuda.ORB.create(
    nfeatures=1500, 
    scaleFactor=1.2,
    nlevels=8,
    edgeThreshold=31,
    firstLevel=0,
    WTA_K=4,
    scoreType=cv2.ORB_HARRIS_SCORE
)

print(f"GPU Image Channels: {gpu_image.channels()}, Type: {gpu_image.type()}")

# Detect and compute descriptors
keypoints, gpu_descriptors = orb.detectAndComputeAsync(gpu_image, None)

The gpu_image I am passing has a size of (180, 290), a channel of 1, and a type of 0

Relevant C++ code referred by the error: The error message points to an assertion failure in the following C++ code within OpenCV's CUDA implementation:

cv::cuda::GpuMat::GpuMat(const GpuMat& m, Rect roi) :
    flags(m.flags), rows(roi.height), cols(roi.width),
    step(m.step), data(m.data + roi.y * step), refcount(m.refcount),
    datastart(m.datastart), dataend(m.dataend),
    allocator(m.allocator)
{
    data += roi.x * elemSize();

    CV_Assert(0 <= roi.x && 0 <= roi.width &&
              roi.x + roi.width <= m.cols &&
              0 <= roi.y && 0 <= roi.height &&
              roi.y + roi.height <= m.rows);

    if (refcount)
        CV_XADD(refcount, 1);

    if (rows <= 0 || cols <= 0)
        rows = cols = 0;
    updateContinuityFlag();
}

Additional Details: Environment: OpenCV 4.10.0-dev with CUDA support. GPU Image Size: (180, 290) (Height x Width). Image Type: Single-channel (grayscale) with type 0

What I've tried: I modified the code to ensure that the ROI is within the bounds of the GPU image, but I still get the error.

Any guidance on what could be causing this issue or how to fix it would be greatly appreciated.

本文标签： opencvAssertion failed in ORB CUDA Error with GpuMat and ROI in detectAndComputeAsyncStack Overflow