Embedded system for IA Assistant

docker
raspberrypi3

#1

Hello,

I got a large project from school, the goal of this project is to develop a connected device like a Google Home but a lot more simple and for kids usage :), We have a lot of libraries to install on our linux software and it’s very annoying to install libraries manually or with sh script on Raspberry pi. So, Docker seemed nice to use. Creating image with everything installed on it and launch the container, like a dump ready to use. But we need audio, video back and also the return of sensor data.
We tried docker on normal processor no rapsberry but we got already an issue for launching video, we use Opencv as library to read video.

So, I discovered Resin.io , do you think Resin can help me to implement all the sensor ?
The Resin.io dump can help me to read video with this libraries ?

Thanks in advance for your responses.


#2

There are people building video based products using resin and the Raspberry Pi. One example is:

I’m not sure if there is an example using OpenCV but it should certainly be possible.


#3

Thanks for your answer !!!
I will try it , tonight :slight_smile:

Thanks :slight_smile:


#4

With resinOS you’re working within a Docker container so you would be able to choose a base image such as e.g. Debian/Raspbian or Arch or Alpine which come with pre-built versions of OpenCV.


#5

so I can use https://hub.docker.com/r/resin/rpi-raspbian/ right ?

To access the Dockerfile at the root of the project ,We have to deploy first and then access the dockerfile to change the base image ? or we can push our dockerfile when we deploy ?


#6

Using the resin-cam as an example:

The FROM line selects the Debian Jessie distribution, then subsequent lines allow us to install some packages we need using APT (in your case this might be OpenCV packages).

There are more docs here:

https://docs.resin.io/raspberrypi3/python/getting-started/

Specifically for how deployment works:

https://docs.resin.io/raspberrypi3/python/getting-started/#deploying-code