LoG with PyTorch

src.pytorch_log.benchmark_log(num_runs=1)

Compare different LoG from Sklearn with PyTorch implementation

Parameters

num_runs (int) – number of runs

src.pytorch_log.log_kernel(kernel_size, sigma_start=1.0, sigma_stop=10.0, num_sigma_steps=10, lin_sigma=True)

Creates Kernels used for Laplacian of Gaussian convolutions

Parameters

  • kernel_size (int) – size of the kernel, should be odd

  • sigma_start (float) – smallest kernel parameter

  • sigma_stop (float) – biggest kernel parameter

  • num_sigma_steps (int) – number of different kernels

  • lin_sigma (bool) – use linear or logarithmic scaling between sigma_start and sigma_end

Returns

kernel

src.pytorch_log.pt_log(img, min_sigma=5, max_sigma=5, num_sigma=1, exclude_borders=True)

PyTorch implementation of Laplassian of Gaussian

Parameters

  • img – input image

  • sigma_start (float) – smallest kernel parameter

  • sigma_stop (float) – biggest kernel parameter

  • num_sigma_steps (int) – number of different kernels

  • exclude_borders (bool) – exclude extrema close to the image border

Returns

x-location, y-location, radius of blobs