Function Converges into a Black Background Picture

[gmh1033286494]

im1 = fixed_path ,im2 = moving_path ,fix_rotation = False,output_dir='/home/chixing/reg_source/reg/',gpu_device = 0,transform_type = 'similarity',lr = 0.01,iterations =10000,min_object_size = 500,no_segment_analysis = True,black_background = True

The wraped moving image become a total black picture.

And when I try Ir = 0.1, iterations = 1000, I got the same result.

[jlevy44]

Hey @gmh1033286494 I can't quite see your picture here, can you upload again?

[gmh1033286494]

[jlevy44]

this is quite strange... you may want to consider manipulating the image, maybe make the background white and refit the model and then transform the original image using the calculated displacement field... there is also an option to input an image mask so it knows which pixels to compute the registration loss.