liu-zhy / temporal-adaptive-module

TAM: Temporal Adaptive Module for Video Recognition

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

About n_segment

Kimsure opened this issue · comments

n_segment是视频序列中帧的个数,但如果不确定视频序列帧数应该怎么办呢,这里不能自适应调整嘛?

设置好n_segment,在训练加载数据时,别管你有多少帧,都会去里边取n_segment帧,所以理论上训练时都是按照n_segment帧训练

commented

In practice, to solve this issue, you can simply replicate the last frame to obtain n-frame input.