o oi8 @sJddlmZmZddlZddlmZmZmZddlm Z ddl m Z m Z de defdd Z d(d ed eee fdedeedef ddZd)d ededeedefddZd)d ededeedefddZd edefddZd*d ede de defddZGdddeZGd d!d!eZGd"d#d#eZGd$d%d%eZGd&d'd'eZdS)+)OptionalUnionN)ModuleTensortensorKORNIA_CHECK_SHAPE)gaussian_blur2dspatial_gradientsigmareturncCs(td|d}|ddkr|d7}|S)Ng @?r)int)r ksizerL/home/ubuntu/.local/lib/python3.10/site-packages/kornia/feature/responses.py_get_kernel_sizes r{Gz?sobelinputk grads_modesigmasc Cst|gd|dur2t|tstdt|t|jdkr*|d|dkr2td|jt ||}|dddddf}|dddddf}t |ddd }t |ddd }t ||dd } ||| | } ||} | || d} |dur| | d  d ddd} | S) uYCompute the Harris cornerness function. .. image:: _static/img/harris_response.png Function does not do any normalization or nms. The response map is computed according the following formulation: .. math:: R = max(0, det(M) - k \cdot trace(M)^2) where: .. math:: M = \sum_{(x,y) \in W} \begin{bmatrix} I^{2}_x & I_x I_y \\ I_x I_y & I^{2}_y \\ \end{bmatrix} and :math:`k` is an empirically determined constant :math:`k ∈ [ 0.04 , 0.06 ]` Args: input: input image with shape :math:`(B, C, H, W)`. k: the Harris detector free parameter. grads_mode: can be ``'sobel'`` for standalone use or ``'diff'`` for use on Gaussian pyramid. sigmas: coefficients to be multiplied by multichannel response. Should be shape of :math:`(B)` It is necessary for performing non-maxima-suppression across different scale pyramid levels. See `vlfeat `_. Return: the response map per channel with shape :math:`(B, C, H, W)`. Example: >>> input = torch.tensor([[[ ... [0., 0., 0., 0., 0., 0., 0.], ... [0., 1., 1., 1., 1., 1., 0.], ... [0., 1., 1., 1., 1., 1., 0.], ... [0., 1., 1., 1., 1., 1., 0.], ... [0., 1., 1., 1., 1., 1., 0.], ... [0., 1., 1., 1., 1., 1., 0.], ... [0., 0., 0., 0., 0., 0., 0.], ... ]]]) # 1x1x7x7 >>> # compute the response map harris_response(input, 0.04) tensor([[[[0.0012, 0.0039, 0.0020, 0.0000, 0.0020, 0.0039, 0.0012], [0.0039, 0.0065, 0.0040, 0.0000, 0.0040, 0.0065, 0.0039], [0.0020, 0.0040, 0.0029, 0.0000, 0.0029, 0.0040, 0.0020], [0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000], [0.0020, 0.0040, 0.0029, 0.0000, 0.0029, 0.0040, 0.0020], [0.0039, 0.0065, 0.0040, 0.0000, 0.0040, 0.0065, 0.0039], [0.0012, 0.0039, 0.0020, 0.0000, 0.0020, 0.0039, 0.0012]]]]) BCHWN!sigmas type is not a Tensor. Got rr9Invalid sigmas shape, we expect B == input.size(0). Got: rr#r r ) r isinstancer TypeErrortypelenshapesize ValueErrorr r powview) rrrr gradientsdxdydx2dy2dxydet_mtrace_mscoresrrrharris_response's$9 " r9cCst|gdt||}|dddddf}|dddddf}t|ddd}t|ddd}t||dd}||||} ||} d| t| dd | } d| t| dd | } t| | } |dur| |d d ddd} | S) aCompute the Shi-Tomasi cornerness function. .. image:: _static/img/gftt_response.png Function does not do any normalization or nms. The response map is computed according the following formulation: .. math:: R = min(eig(M)) where: .. math:: M = \sum_{(x,y) \in W} \begin{bmatrix} I^{2}_x & I_x I_y \\ I_x I_y & I^{2}_y \\ \end{bmatrix} Args: input: input image with shape :math:`(B, C, H, W)`. grads_mode: can be ``'sobel'`` for standalone use or ``'diff'`` for use on Gaussian pyramid. sigmas: coefficients to be multiplied by multichannel response. Should be shape of :math:`(B)` It is necessary for performing non-maxima-suppression across different scale pyramid levels. See `vlfeat `_. Return: the response map per channel with shape :math:`(B, C, H, W)`. Example: >>> input = torch.tensor([[[ ... [0., 0., 0., 0., 0., 0., 0.], ... [0., 1., 1., 1., 1., 1., 0.], ... [0., 1., 1., 1., 1., 1., 0.], ... [0., 1., 1., 1., 1., 1., 0.], ... [0., 1., 1., 1., 1., 1., 0.], ... [0., 1., 1., 1., 1., 1., 0.], ... [0., 0., 0., 0., 0., 0., 0.], ... ]]]) # 1x1x7x7 >>> # compute the response map gftt_response(input) tensor([[[[0.0155, 0.0334, 0.0194, 0.0000, 0.0194, 0.0334, 0.0155], [0.0334, 0.0575, 0.0339, 0.0000, 0.0339, 0.0575, 0.0334], [0.0194, 0.0339, 0.0497, 0.0000, 0.0497, 0.0339, 0.0194], [0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000], [0.0194, 0.0339, 0.0497, 0.0000, 0.0497, 0.0339, 0.0194], [0.0334, 0.0575, 0.0339, 0.0000, 0.0339, 0.0575, 0.0334], [0.0155, 0.0334, 0.0194, 0.0000, 0.0194, 0.0334, 0.0155]]]]) rNrrrr"r$g?r%r&) rr r torchsqrtabsminr.r/)rrrr0r1r2r3r4r5r6r7e1e2r8rrr gftt_response~s3 "" r@cCst|gd|dur2t|tstdt|t|jdkr*|d|dkr2td|jt ||d}|dddddf}|dddddf}|dddddf}|||d}|durr|| d d ddd}|S) aCompute the absolute of determinant of the Hessian matrix. .. image:: _static/img/hessian_response.png Function does not do any normalization or nms. The response map is computed according the following formulation: .. math:: R = det(H) where: .. math:: M = \sum_{(x,y) \in W} \begin{bmatrix} I_{xx} & I_{xy} \\ I_{xy} & I_{yy} \\ \end{bmatrix} Args: input: input image with shape :math:`(B, C, H, W)`. grads_mode: can be ``'sobel'`` for standalone use or ``'diff'`` for use on Gaussian pyramid. sigmas: coefficients to be multiplied by multichannel response. Should be shape of :math:`(B)` It is necessary for performing non-maxima-suppression across different scale pyramid levels. See `vlfeat `_. Return: the response map per channel with shape :math:`(B, C, H, W)`. Shape: - Input: :math:`(B, C, H, W)` - Output: :math:`(B, C, H, W)` Examples: >>> input = torch.tensor([[[ ... [0., 0., 0., 0., 0., 0., 0.], ... [0., 1., 1., 1., 1., 1., 0.], ... [0., 1., 1., 1., 1., 1., 0.], ... [0., 1., 1., 1., 1., 1., 0.], ... [0., 1., 1., 1., 1., 1., 0.], ... [0., 1., 1., 1., 1., 1., 0.], ... [0., 0., 0., 0., 0., 0., 0.], ... ]]]) # 1x1x7x7 >>> # compute the response map hessian_response(input) tensor([[[[0.0155, 0.0334, 0.0194, 0.0000, 0.0194, 0.0334, 0.0155], [0.0334, 0.0575, 0.0339, 0.0000, 0.0339, 0.0575, 0.0334], [0.0194, 0.0339, 0.0497, 0.0000, 0.0497, 0.0339, 0.0194], [0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000], [0.0194, 0.0339, 0.0497, 0.0000, 0.0497, 0.0339, 0.0194], [0.0334, 0.0575, 0.0339, 0.0000, 0.0339, 0.0575, 0.0334], [0.0155, 0.0334, 0.0194, 0.0000, 0.0194, 0.0334, 0.0155]]]]) rNr rrr!rr%r&) rr'rr(r)r*r+r,r-r r.r/)rrrr0dxxr5dyyr8rrrhessian_responses7 " rCcCsBt|gd|ddddddf|ddddddfS)zCompute the Difference-of-Gaussian response. Args: input: a given the gaussian 5d tensor :math:`(B, C, D, H, W)`. Return: the response map per channel with shape :math:`(B, C, D-1, H, W)`. )rrLrrNrr&r)rrrr dog_responses 4rEr 皙?sigma1sigma2cCsNt|gdt|}t|}t|||f||f}t|||f||f}||S)aSCompute the Difference-of-Gaussian response. .. image:: _static/img/dog_response_single.png Args: input: a given the gaussian 4d tensor :math:`(B, C, H, W)`. sigma1: lower gaussian sigma sigma2: bigger gaussian sigma Return: the response map per channel with shape :math:`(B, C, H, W)`. r)rrr )rrGrHks1ks2g1g2rrrdog_response_single%s rMcsLeZdZdZd fdd ZdefddZd ded eedefd d Z Z S)BlobDoGzzModule that calculates Difference-of-Gaussians blobs. See :func: `~kornia.feature.dog_response` for details. r NcstdSN)super__init__self __class__rrrQBzBlobDoG.__init__cCs|jjSrO)rU__name__rRrrr__repr__EzBlobDoG.__repr__rrcCst|SrO)rErSrrrrrforwardHrYzBlobDoG.forward)r NrO) rW __module__ __qualname____doc__rQstrrXrrr[ __classcell__rrrTrrN;s $rNcsZeZdZdZddededdffdd Zdefd d Zdd ed e edefddZ Z S) BlobDoGSinglezModule that calculates Difference-of-Gaussians blobs. .. image:: _static/img/dog_response_single.png See :func:`~kornia.feature.dog_response_single` for details. r rFrGrHr Ncst||_||_dSrO)rPrQrGrH)rSrGrHrTrrrQTs  zBlobDoGSingle.__init__cC|jjd|jd|jdS)Nz , sigma1=z , sigma2=))rUrWrGrHrRrrrrXYzBlobDoGSingle.__repr__rrcCst||j|jSrO)rMrGrHrZrrrr[\szBlobDoGSingle.forwardr rFrO) rWr\r]r^floatrQr_rXrrr[r`rrrTrraLs $racsleZdZUdZeed<ddeeefdeddffdd Z defd d Z dd ed e edefd dZ Z S) CornerHarriszModule that calculates Harris corners. .. image:: _static/img/harris_response.png See :func:`~kornia.feature.harris_response` for details. rrrr Ncs<tt|tr|dt|n|d|||_dS)Nr)rPrQr'rfregister_bufferrr)rSrrrTrrrQjs    zCornerHarris.__init__cCrb)Nz(k=z , grads_mode=rc)rUrWrrrRrrrrXrrdzCornerHarris.__repr__rrcCst||j|j|SrO)r9rrrZrrrr[uszCornerHarris.forwardrrO)rWr\r]r^r__annotations__rrfr_rQrXrr[r`rrrTrrg`s $$rgcVeZdZdZddeddffdd Zdefdd Zdd ed eedefd d Z Z S) CornerGFTTzModule that calculates Shi-Tomasi corners. .. image:: _static/img/gftt_response.png See :func:`~kornia.feature.gftt_response` for details. rrr Nct||_dSrOrPrQrrSrrTrrrQ  zCornerGFTT.__init__cC|jjd|jdSNz (grads_mode=rcrUrWrrRrrrrXzCornerGFTT.__repr__rrcCt||j|SrO)r@rrZrrrr[rVzCornerGFTT.forwardrirO rWr\r]r^r_rQrXrrr[r`rrrTrrly $rlcrk) BlobHessianzModule that calculates Hessian blobs. .. image:: _static/img/hessian_response.png See :func:`~kornia.feature.hessian_response` for details. rrr NcrmrOrnrorTrrrQrpzBlobHessian.__init__cCrqrrrsrRrrrrXrtzBlobHessian.__repr__rrcCrurO)rCrrZrrrr[rVzBlobHessian.forwardrirOrvrrrTrrxrwrx)rrN)rNre)typingrrr: kornia.corerrrkornia.core.checkrkornia.filtersr r rfrrr_r9r@rCrErMrNrargrlrxrrrrs6    W KM