Is there any mistake in the top-down densely connections ?

[TooPigToBig]

In the GFF paper, top-down densely connections are shown as the picture

And I think the code should be as follows d5 = tf.concat([psp ], axis=-1) d4 = tf.concat([psp, x4gff], axis=-1) d3 = tf.concat([psp, x3gff, x4gff], axis=-1) d2 = tf.concat([psp, x2gff, x3gff, x4gff], axis=-1) d1 = tf.concat([psp, x1gff, x2gff, x3gff, x4gff], axis=-1) full_block = tf.concat([d1, d2, d3, d4, d5], axis=-1)

howerer in you code the top-down densely connections are as follows

Could you please answer my confusion？

[nhatuan84]

Sorry long time i do not remember. But you can try ur code. I think ur code also converge.

On Sun, 23 Feb 2020, 19:48 hxtisme, [email protected] wrote:

In the GFF paper, top-down densely connections are shown as the picture [image: image] https://user-images.githubusercontent.com/53428507/75112160-fafbe600-567b-11ea-8c2f-cc3ce81d6979.png

And I think the code should be as follows d5 = tf.concat([psp, ], axis=-1) d4 = tf.concat([psp, , x4gff], axis=-1) d3 = tf.concat([psp, , x3gff, x4gff], axis=-1) d2 = tf.concat([psp, x2gff, x3gff, x4gff], axis=-1) d1 = tf.concat([psp, x1gff, x2gff, x3gff, x4gff], axis=-1) full_block = tf.concat([d1, d2, d3, d4, d5], axis=-1)

howerer in you code the top-down densely connections are as follows [image: image] https://user-images.githubusercontent.com/53428507/75112152-e3246200-567b-11ea-9026-16ec839c53b1.png

Could you please answer my confusion？

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/nhatuan84/GFF-Gated-Fully-Fusion-for-Semantic-Segmentation/issues/1?email_source=notifications&email_token=ABEN6LWJ443UBSYK73A5XLTREJWCHA5CNFSM4KZ2KUN2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4IPRXLYA, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABEN6LV2QVZHHNPZNRT2EMDREJWCHANCNFSM4KZ2KUNQ .