update
This commit is contained in:
@@ -92,6 +92,8 @@ SSD is a multi-resolution object detection
|
||||
|
||||
RetinaNet combine feature pyramid network with focal loss to reduce the standard cross-entropy loss for well-classified examples.
|
||||
|
||||

|
||||
|
||||
> Cross-entropy loss:
|
||||
> $$CE(p_t) = - \log(p_t)$$
|
||||
|
||||
|
||||
Reference in New Issue
Block a user