Is Computer Vision viable for Quality Control?
retained text
I am starting a factory where I will be running Injection Moulding machines which take at least 10-20 seconds and upwards of minutes to produce one part. I am looking for QC methods for production parts since I will be dealing with big sized corporations and they do not play with quality control haha. I have my bachelors in CompSci, and did explore CV a little bit during college but did not get far into it. Now while I was thinking of ways for improving the manufacturing unit, QC using CV came to my mind. I do see it being possible from other's experiences online and from an amateur perspective it does make sense. My goal is to build something that would be good for a go/no go system. I do not need to see super detailed results (for now) regarding the precision and dimensions, it is more about "will this work for someone who is inspecting it with the naked eye?" I do see a pretty big challenge right out of the gate - training data. I don't think I will find a big enough dataset of the parts that I manufacture to train the model. I can take images of those parts manually, since I do get them made from outside and sell them currently, this is for a switch to an in house manufacturing setup. But that would no doubt take a LONG TIME to get a decent sized data set. I will look into it more if there is something I can use, but would appreciate suggestions on that. I also hear some people mention lighting being an issue. Would appreciate any insights on that. Also would appreciate if I could get some cost details regarding the setup, like what equipment I will need. I can run training on my own machines that I have, they are fairly powerful, I am asking primarily for inference and rejection/acceptance. I will start learning about it more, the topic is fun, and see its feasibility in multiple aspects with my work. Appreciate all the feedback and help anyone has. Thank you very much!
document fetched 2026-09-03 00:24 · canonical https://reddit.com/r/computervision/comments/1w59ibn/is_computer_vision_viable_for_quality_control