Install checkup for requirements - Suggestion/Improvement - Githubissues

SharpAI / DeepCamera

Open-Source AI Camera. Empower any camera/CCTV with state-of-the-art AI, including facial recognition, person recognition(RE-ID) car detection, fall detection and more

https://sharpai.github.io/DeepCamera/

MIT License

1.82k stars 311 forks source link

Install checkup for requirements - Suggestion/Improvement #65

Closed staberas closed 2 years ago

staberas commented 2 years ago

Enviroment : Jetson nano - 2GB version

The ./run-on-nano.sh script requires a check if the user has :

Has installed the correct version of docker-compose (1.27.4) , i had the 1.17 version and had to manually find from docker website the correct version, i also had problem with pip so a check up if docker is correctly installed before it executes the commands would be nice.
After running ./run-on-nano.sh start , jetson stopped responding since its 29GB memory got filled up, a check for sufficient storage before installing would be nice

Thanks for the great job you did i hope you implement my suggestions/improvements

solderzzc commented 2 years ago

Thanks for the advise, I'll add the env checks.

For issue 2, do you have 4G memory version of Nano for testing, I haven't test on 2GB memory version of Nano yet. I think the 29GB storage is big enough.

staberas commented 2 years ago

How much hard disk space does docker requires? (my jetson disk space is actually filled with other vision projects) i have the 2gb ram version but expanded with swap memory

solderzzc commented 2 years ago

@staberas I'll checkup the benchmark of 2gb Nano since I have one but haven't run deepcamera on it. If it fails, I'll try to optimize it later. For disk space, it should be something like less than 4-6GB.

staberas commented 2 years ago

Managed to install and it just about works, its freezes and occupies 80-90% of memory which triggers swapping, this means it kills the vnc viewer and you can only control it via ssh. Other than that shinobi is accesible on the network though a bit slow to load.

For my application i just need the serverAPI ( ./start-cloud.sh ) for face recognition , since i have a sipeed M1N to do the face detection and cropping for the serverAPI to learn and recognise. Im i correct ?

solderzzc commented 2 years ago

API server provides label interface and management GUI for devices, face recognition doesn't run on API Server. Face recognition is running on device, it opens port 3000 to accept image to be processed.

For your reference: https://github.com/SharpAI/DeepCamera/blob/nano/src/detector/index.js#L917

solderzzc commented 2 years ago

Following is the Android Java code to upload face image and build json message then send to DeepCamera for face recognition: https://github.com/SharpAI/AI_Camera/blob/master/sharpaiLib/src/main/java/org/sharpai/lib/Detection.java#L657

solderzzc commented 2 years ago

Release v3.4.3 addressed installation checkup: https://github.com/SharpAI/DeepCamera/releases/tag/v3.4.3

solderzzc commented 2 years ago

Fixed it with v3.4.3 release. @staberas thanks for your advise, we released a cli command line which can be easily installed by pip install, then one command can start the service instead of start with 'git clone'.

sharpai-cli deepcamera start