Open jiandai opened 7 years ago
In your unet architecture, "UpSampling2D" is used for "deconvolution". I saw in this package https://github.com/jakeret/tf_unet the author uses transposed convolution. Does this choice have any impact on the performance?
In your unet architecture, "UpSampling2D" is used for "deconvolution". I saw in this package https://github.com/jakeret/tf_unet the author uses transposed convolution. Does this choice have any impact on the performance?