search

aim-uofa / AdelaiDet

AdelaiDet is an open source toolbox for multiple instance-level detection and recognition tasks.
https://git.io/AdelaiDet
Other
3.37k stars 646 forks source link

About 169 parameters in CondInst #11

Closed sxhxliang closed 4 years ago

sxhxliang commented 4 years ago

Conditional Convolutions for Instance Segmentation brings a new paradigm to instance segmentation, but I don't quite understand some of the details in the paper. The following three questions are my doubts. Can you give me some advice?

  1. Are all masks calculated from the positive samples generated by FCOS? (subsection 2.2. Network Outputs and Training Targets)
  2. Why are there 169 parameters ( vector ) in the Controller Head and How do the 169 parameters assign to three 1 × 1 convolutions with 8-channels in Mask FCN Head? How many parameters are there in each of the three 1 × 1 convolution layers?（subsection 2.2.）

  3. How are the three conditional convolution layers and the corresponding parameters(169 parameters in total) calculated in Mask FCN Head? (subsection 2.4)

    Thank you very much for your time.

tianzhi0549 commented 4 years ago

@AaronLeong Thank you for reaching out. 1) Yes. 2) 169 = #weights + #biases. #weights = (8 + 2) 8 (conv1) + 8 8 (conv2) + 8 * 1 (conv3), #biases = 8 + 8 + 1. Note that the conv1 has 10-channel input feature maps because we append the relative coordinates. 3) The 169 parameters are predicted by the controller head, which is why we call the method conditional convolutions.

sxhxliang commented 4 years ago

@AaronLeong Thank you for reaching out.

  1. Yes.
  2. 169 = #weights + #biases. #weights = (8 + 2) 8 (conv1) + 8 8 (conv2) + 8 * 1 (conv3), #biases = 8 + 8 + 1. Note that the conv1 has 10-channel input feature maps because we append the relative coordinates.
  3. The 169 parameters are predicted by the controller head, which is why we call the method conditional convolutions.

Thank you, nice work!