Hi everyone 👋,
Welcome to the introduction to what will be a 5 *(-ish)* part series of blog posts regarding the Protostar Vision Box. This is a solution for visual inspection of various goods, powered by AI for near perfect defect detection. It is designed to catch various anomalies on products, such as deformed bottle caps, misaligned screw holes, surface scratches, etc. It can be mounted directly on the conveyor line or mounted as a standalone version. Where might this solution fit in? From quality inspection in the food industry all the way to the manufacturing industry, anywhere where the defects can be seen with the naked eye or, in this case, a camera. This series will cover the hardware and software makeup, ML model development and some insights we gathered along the way. It will consist of the following posts, from the perspective of the quality inspection of a bottling process:
- Hardware overview (lighting, camera) – to have consistent images, we need consistent lighting and proper cameras. This post will go into detail about how we designed the hardware and implemented our custom lighting solution and which cameras we chose.
- Sensors suite – what sensors we are using, why those sensors were chosen, technical specs and what information we can gather from them
- Artificial intelligence – describing the brain of the solution, what needed to be gathered, how it was made and implemented
- Color detection algorithm – or how to use classical development to expand upon the AI model capabilities
- What went wrong – a summary of our challenges, what hurdles we had to cross and everything we learned during this project. Given the sheer number of small details we noticed, this one may not fit in just *one* blog post
And that sums it up, you can expect new posts every Friday (what we’re calling ‘PVB Fridays’ 😀). We’re excited to share this journey with you! Let us know your thoughts at hello@protostar.ai.
