💪 2019 SW Festival Home PT 💪

딥러닝을 활용한 홈짐 트레이닝 서비스

Dongguk Univ BigData Farm

팀원

• 김성진

• 이태민

• 정의정

• 채진영

SW

Purpose

어디서든 편하게 운동할 수 있으며 개개인에게 맞춤형 피드백을 제공하는 딥러닝을 활용한 모바일헬스 서비스를 제작한다.

System Architecture

Features

• Pose Estimate using Object Detection
• Feedback SPEED, POSTURE(knee, wrist)
• Providing cumulative score for increasing Engage

Demo

Demo Command

python main.py

Pose estimate Result

Feedback

Score record

Todo List

• [] sound feedback
• [] Torso normalization
• [] Upgrade UI/UX

Done List

• Train COCO Human data
• Make scoring pose algorithm
• Make base UI/UX
• Denote scores
• Make ground truth skeleton
• Make pose estimation model
• Email to Stanford Univ Student to get dataset

Information

Version

Python : 3.7
Tensorflow : 1.8.0
Opencv : 3.4.2
Pillow : 6.0.0
Pandas : 0.25.0
Matplotlib : 3.1.0
Numpy : 1.16.4
pygame : 1.9.6
scikit-learn : 0.21.3

Weigths Download

The pretrained pose weights file can be downloaded here. Place this weights file under directory ./data/pose_weights/

Reference

• https://github.com/zzh8829/yolov3-tf2

• https://github.com/ZackPashkin/YOLOv3-EfficientNet-EffYolo