search

VITA-Group / FasterSeg

[ICLR 2020] "FasterSeg: Searching for Faster Real-time Semantic Segmentation" by Wuyang Chen, Xinyu Gong, Xianming Liu, Qian Zhang, Yuan Li, Zhangyang Wang
MIT License
525 stars 107 forks source link

Why do you weight latencies from different scales differently? #20

Closed maaft closed 4 years ago

maaft commented 4 years ago

In your code you weight the latency coming from each branch differently with 

scale_latency_weights=[3./12, 4./12, 5./12]

in: https://github.com/TAMU-VITA/FasterSeg/blob/bb52a004ff83f64d3dd8989104234fdb862d1cc5/search/model_search.py#L474

Can you elaborate on why you do this?

chenwydj commented 4 years ago

Hi @maaft ! I just updated the repo. You don't have to use this weight.