Open-Source Smart Cameras
Deploy GreenThumb’s Commercial Computer Vision Models
GreenThumb IO is a cannabis computer vision and AI product. We are open-sourcing our prototypes that we used to develop our platform capable of integrating with any industry standard product and service providers. I have a series of blogs outlining our projects which can be found here on Medium — follow to stay in the loop on new developments!
SmartCam.Dev
Developing smart cameras is something I’ve been passionate about for years. Now having assisted in the design, development, and deployment of commercial grade smart camera technology in agricultural settings, I have experienced the pains of deploying hundreds of cameras in the field.
It’s not only a matter of installing the infrastructure to power and network a CCTV system, but to process all the video feeds in real-time, meaning you can’t rely on streaming to the cloud for Computer Vision base AI model inference.
I will also be launching SmartCam.dev which will host more resources for anyone interested.
Edge Computing
Drones are actually quite customizable by design from hardware to software that is rugged and robust. All Power Over Ethernet POE CCTV cameras have processors much like your iPhone or Android Smart Phone. Both iOS and Android Operating Systems provide SDKs for developers to deploy their trained Deep Learning models to the App Store using iOS’s Core ML or Android’s ML Kit. Smart Cameras like the NestCam iQ have powerful AI features baked in, and do provide an API for developers to retrieve motion captured videos as well as periodic snapshots, but there is no SDK for a developer to release their own custom trained model to their App’s user’s cameras. Because they rely on the cloud, NestCams aren’t a viable solution as they cost too much bandwidth.
Computer Vision-based AI startups are faced with two deployment strategies at the moment: build your own, or deploy servers to process video feeds alongside of legacy dumb POE cameras. This is where what I’ve been working on comes in.
I’ve made the leap and purchased both SmartCam.dev and SmartCamera.dev in an effort to advance the industry by providing an initial commercially viable smart camera. This site will launch within the coming weeks.
Below, this is the camera I’ve developed as V1, and it will be available for purchase as well as instructions on how to build your own. If there is significant interest in an Open Smart Camera SDK, I intend to do a crowdfund to make a product that is both visually appealing and industrially viable.
Hardware & Costs
DIY PoE Deployable Smart Camera —Total Cost: $145
Industrial Weatherproof Aluminum Housing — $27
Raspberry Pi 3 B+ — $40
SD Card — $8
IR Camera Sensor — $30
PoE to USB — $10
Neural Compute Strick VPU (optional) — $75
I’ll be providing an STL of a 3D printed faceplate, pi mount, and poe bracket in good time. I’m also working a designing a more compact and contemporary housing which I’ll also provide design files for in the next few months.
Software
I’m not going to go deep into the software today, but I highly recommend these resources below to get you setup with your Smart Camera project:
Conclusion
This is just getting started, and I plan to expand on this blog, which is more of a blueprint, into a full featured guide. If you’re interested in contributing to the project please contact me at justin@greenthumb.io.
This is only V1, V2 will of the camera will look much more sleek. Follow along here on Medium to stay tuned for more in this series of Open Source Hardware Project Announcements! I am currently building out a side project, smartcam.dev, which will be the home of an online store and guides to buy or build your own deployable smart cameras.
If you have any questions or are interested in the project or using my open-source packages, feel free to leave a comment or ping me on LinkedIn or Twitter, I will be responsive.
GreenThumb is also seeking early clients, partners, and initial funding. Yes, I did the work and paid for this all on my own.
