No bias for each convolutional layer? Why?

msracver / FCIS

Fully Convolutional Instance-aware Semantic Segmentation

MIT License

1.57k stars 415 forks source link

No bias for each convolutional layer? Why? #128

Closed zhaodongsun closed 6 years ago

zhaodongsun commented 6 years ago

At /FCIS/fcis/symbols/resnet_v1_101_fcis.py, I found the no_bias parameter is set True ,which means there is no bias for each convolutional layer. Why do you remove the bias?

HaozhiQi commented 6 years ago

Biases are in the BN layers that follow.