How to get the inverse deform filed?

[1190202328]

I am a beginner and I would like to know how TransMorph-diff obtains an inverse deformation field? Can I use this method below?

flow_mean = self.flow_mean(x)
log_sigma = self.flow_sigma(x)
noise = torch.randn(flow_mean.shape).cuda()
if self.training:
    flow = flow_mean + torch.exp(log_sigma / 2.0) * noise
else:
    flow = flow_mean  # + 1 * noise

# get inverse
inverse_flow = -flow

for _ in range(self.int_steps):
    deform_field = flow + self.low_res_id_transform
    flow_1 = self.bilinear(flow, deform_field)
    flow = flow_1 + flow
disp_field = F.interpolate(flow, scale_factor=self.sample_rate, mode='bilinear')
disp_field = self.scale_map(disp_field, np.array([1, 1]))
deform_field = disp_field + affine_map
warped_source = self.bilinear_img(source_image, deform_field)
self.res_flow_mean = flow
self.res_log_sigma = log_sigma

for _ in range(self.int_steps):
    inverse_deform_field = inverse_flow + self.low_res_id_transform
    inverse_flow_1 = self.bilinear(inverse_flow, inverse_deform_field)
    inverse_flow = inverse_flow_1 + inverse_flow
inverse_disp_field = F.interpolate(inverse_flow, scale_factor=self.sample_rate, mode='bilinear')
inverse_disp_field = self.scale_map(inverse_disp_field, np.array([1, 1]))
inverse_deform_field = inverse_disp_field + affine_map
warped_target = self.bilinear_img(target_image, inverse_deform_field)
self.inverse_res_flow_mean = inverse_flow

Thanks!