A pytorch-version implementation codes of paper: "BMN: Boundary-Matching Network for Temporal Action Proposal Generation", which is accepted in ICCV 2019.
MIT License
291
stars
64
forks
source link
There is a gap of mAp between reproduced result and original paper #46
Hi,I use the proposal result of your code, and try to reproduce the detection result follow the original paper. But the mAp result has a gap between original paper. Did you have this problem, and how did you solve that. Thanks!
Hi,I use the proposal result of your code, and try to reproduce the detection result follow the original paper. But the mAp result has a gap between original paper. Did you have this problem, and how did you solve that. Thanks!