def forward(ctx, x, angles, tex_cache, rays_cfg):
sinogram = torch_radon_cuda.forward(x, angles, tex_cache, rays_cfg)
ctx.tex_cache = tex_cache
ctx.rays_cfg = rays_cfg
ctx.save_for_backward(angles)
return sinogram
def backward(ctx, grad_x):
if not grad_x.is_contiguous():
grad_x = grad_x.contiguous()
angles, = ctx.saved_variables
grad = torch_radon_cuda.forward(grad_x, angles, ctx.tex_cache,
ctx.rays_cfg)
return grad, None, None, None
def forward(ctx, x, det_count, det_spacing, angles, tex_cache, clip_to_circle):
sinogram = torch_radon_cuda.forward(x, det_count, det_spacing, angles, tex_cache, clip_to_circle)
ctx.tex_cache = tex_cache
ctx.det_count = det_count
ctx.det_spacing = det_spacing
ctx.clip_to_circle = clip_to_circle
ctx.save_for_backward(angles)
return sinogram
def backward(ctx, grad_x):
angles, = ctx.saved_variables
grad = torch_radon_cuda.forward(grad_x, ctx.det_count, ctx.det_spacing, angles, ctx.tex_cache,
ctx.clip_to_circle)
return grad, None, None, None, None, None
def backward(ctx, grad_x):
angles, = ctx.saved_variables
grad = torch_radon_cuda.forward(grad_x, angles, ctx.tex_cache,
ctx.rays_cfg)
return grad, None, None, None