-
Notifications
You must be signed in to change notification settings - Fork 2
[FR] scan with camera #15
Comments
Thanks for the idea. Unfortunately I'm not sure how doable it will be given how small resistors are. And also the 5 band blue bodied resistors don't have much contrast. |
This is a very good idea, but I think the first step would be to collect data. A while ago I tried some basic tests just looking at the value of pixels (after auto white balance) and found, for the first batch of resistors I tried that colors were easily distinguishable. Without even resorting to machine learning, I could have made a simple heuristic using hue, saturation, and value. However, it broke down when I tried blue bodied resistors. I didn't have enough data to know if that was due to this particular manufacturer's paint. Or is resistor code paint semi-transparentand I'd need to subtract the background? Or do all blue resistors use those slightly different paint colors? And what happens when the pictures are taken with different cameras, in different lighting conditions? I'd like to propose adding an optional feature allowing users, after identifying a resistor manually, to take a magnified photo using the app and upload it to this project. That way there'd be a database of known real images with ground-truth identification. We could then try some of the basic pattern recognition algorithms and see what's possible. I know setting up an online repository and writing camera access code may not be easy. But is it something the developers might consider as beneficial to this project? Feel free to let me know if this suggestion is uninteresting. |
Hi @hackerb9, thanks for the suggestions, but this isn't really something I'm interested in working on right now. Maybe if in future I'm bored I'll think about implementing this again, but storing images tends to be quite expensive also. BTW in my stocks of blue bodied resistors, I have found at least 2 different base shades (lighter and darker background), and the darker background ones are quite hard even for my eyes to identify. |
We must have the same stock of illegible-to-humans resistors! I had to use my phone's magnifying glass app (with light) to decipher them, which is what made me think of automating it. Thanks for considering the idea! |
would be awesome to just "look" at a resistor with camera and recognize the colors.
The text was updated successfully, but these errors were encountered: