Kernel Head in RTMDET
#10032
Replies: 1 comment 1 reply
-
|
@GTrui6 Its a reference from condinst, you can see the original paper |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
Thanks for your reply, i have saw the paper |
Beta Was this translation helpful? Give feedback.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Why the kernel head in RTMDET has 169 dimensions? The paper says:"The vector is divided into three parts (lengths are 88, 72, and 9 respectively), which are used to form the kernels of three dynamic convolution layers". So why is that?
Beta Was this translation helpful? Give feedback.
All reactions