Open dingfg opened 7 years ago
Richer Convolutional Features for Edge Detection: "The hyper-parameter λ is to balance positive and negative samples." I wonder how to compute λ.
Hi. λ is a constant, and its values for different datasets can be found in the paper.
Richer Convolutional Features for Edge Detection: "The hyper-parameter λ is to balance positive and negative samples." I wonder how to compute λ.