HttpInference is a http android client and http server example for deep learning offloading.
- git clone https://github.com/lovgrammer/HttpInference.git
- import project to your Android Studio
- build and install
- Install python3
- ./install.sh
- cd infserver
- python3 manage.py migrate
- python3 manage.py runserver ip:port
- Type in the remote1 and remote2 server IP "http://yourip:yourport/" (you should put / in the end of the string) (e.g., remote1 - edge server, remote2 - cloud server)
- Click start button
- Then, HttpInference app execute three process for 100 images
- run segmentation inference in the local device
- send file to remote1 inference server and get result image from the server
- send file to remote2 inference server and get result image from the server
- The measurement result of latency and segmentation results will be stored in the /sdcard/result.csv
- adb pull /sdcard/result.csv .
- python analysis_inftime.py
- python analysis_segnum.py
- python analysis_totaltime.py
