info.yaml 1.3 KB
Newer Older
H
huangjun12 已提交
1 2 3 4 5
---
Model_Info:
  name: "PP-TSM"
  description: "PP-TSM是PaddleVideo自研的实用产业级视频分类模型,考虑精度和速度的平衡,进行模型瘦身和精度优化,在Kinetics-400数据集上精度较原论文提升4个点。"
  description_en: "PP-TSM is a practical industrial video classification model developed by PaddleVideo. We slim the model size and optimize the accuracy with the considerations of the trade-off between speed and precision and achieve 76.16% in Kinetics-400 dataset."
Q
qiceng 已提交
6
  icon: "https://user-images.githubusercontent.com/59186797/203051256-8ba37e13-a451-41b8-8066-d1e20ec1f621.jpg"
H
huangjun12 已提交
7 8 9 10 11 12 13
  from_repo: "PaddleVideo"
Task:
- tag_en: "Video"
  tag: "视频"
  sub_tag_en: "Short Video Content Classification"
  sub_tag: "短视频内容分类"
Example:
L
liuTINA0907 已提交
14 15
- tag: "互联网"
  tag_en: "Internet"
H
huangjun12 已提交
16 17 18 19 20 21 22
  sub_tag: "时刻剪辑"
  sub_tag_en: "Moment editing"
  title: "基于PaddleVideo的足球精彩时刻剪辑"
  title_en: "Soccer video highlights clip based on PaddleVideo"
  url: "https://aistudio.baidu.com/aistudio/projectdetail/3473391?contributionType=1"
  url_en: 
Datasets: "Kinetics-400, UCF-101"
Q
qiceng 已提交
23
Publisher: "Baidu"
H
huangjun12 已提交
24 25 26 27 28 29
License: "apache.2.0"
Paper:
- title: "TSM: Temporal Shift Module for Efficient Video Understanding"
  url: "https://arxiv.org/pdf/1811.08383.pdf"
IfTraining: 1
IfOnlineDemo: 1