基于AWS DeepLens的姿态识别

# AWS-DeepLens-Pose ## AWS 딥렌즈를 이용한 자세 인식 <p align=center><img width="200" src="https://user-images.githubusercontent.com/68395698/113527692-47448400-95f9-11eb-844d-408fbf3f25e8.png"></p> ## Environment - AWS DeepLens v1.1 - python 3.7 - mxnet - ubuntu ## Train 1. You can train any posture 2. Extract posture data using [extract_pose.py](https://github.com/seoh02h/AWS-DeepLens-Pose/blob/master/extract_pose.py) 3. Train using [mxnet_train.ipynb](https://github.com/seoh02h/AWS-DeepLens-Pose/blob/master/mxnet_train.ipynb) 4. Get 'DeepLens_pose-0500.params' and 'DeepLens_pose-symbol.json' ## Openpose [CMU-Perceptual-Computing-Lab/openpose](https://github.com/CMU-Perceptual-Computing-Lab/openpose) [mxnet_Realtime_Multi-Person_Pose_Estimation](https://github.com/dragonfly90/mxnet_Realtime_Multi-Person_Pose_Estimation) 1. Convert openpose model to run on DeepLens 2. Get 'realtimePose.bin' and 'realtimePose.xml' ## Deploy Reference : [deeplens-inference-lambda-create](https://github.com/awsdocs/aws-deeplens-user-guide/blob/master/doc_source/deeplens-inference-lambda-create.md) 1. Use AWS Lambda ## Project ### Live ``` python deeplens_pose_live.py ``` ### Test with Images ``` python deeplens_pose_image.py ``` ### Result Sample ( Test ) <p align=center><img width="800" src="./result/test1.png"></p> <p align=center><img width="800" src="./result/test2.png"></p> ## View ### Project Streams ``` ./view_deeplens_project_streams.sh ``` ### Live Streams ``` ./view_deeplens_live_streams.sh ```