|
def semantic_inference(self, mask_cls, mask_pred): |
semseg = mask_cls.softmax(-1) @ mask_pred.sigmoid()
which will result in:
1>= semseg >= 0
When we compute
loss_func, we set semseg as a
logits, and compute prob by semseg.sigmoid()
If
1>= semseg >= 0，the probs must be >= 0.5, but we need the prob >=0.
Maybe I am wrong or I miss some calculation process，please feel free to advise me。
SegVit/decode_heads/atm_single_head.py
Line 277 in 8561638
semseg = mask_cls.softmax(-1) @ mask_pred.sigmoid()
which will result in: 1>= semseg >= 0
When we compute loss_func, we set semseg as a logits, and compute prob by semseg.sigmoid()
If 1>= semseg >= 0，the probs must be >= 0.5, but we need the prob >=0.
Maybe I am wrong or I miss some calculation process，please feel free to advise me。