Good Morning from my Robotics Lab! This is Shadow_8472, and today, I’m recounting the tale of my first working prototype, and how I ruined it before getting it to actually work. Let’s get started!
Pi4 8GB and Cards
I am now the owner of a Pi4 with 8GB of memory, the highest end Pi available at the present time. When I unboxed it, I put it directly into a case with a fan powered by GPIO pins. Some day, I’ll want to benchmark its cooling against my Pi 4 with the huge heatsink case and passive airflow.
The cards I had on order never came in, and their listing vanished in the meantime. I ended up with some 64GB cards from Gigastone that are supposed to be better on paper, but I’m not in a position to benchmark the Raspberry Pi use case. While these new cards only have one SD adapter between them –I’ve been using SD adapters for labels– they did include a USB adapter. It’s only 2.0, though.
Manjaro ARM
I have not fully settled on what distro to go with for this project. TinyCore is great for starting projects, but I have a hard time finishing it there. For the time being, I’ll be prototyping from my Manjaro ARM card. Whenever I need to reset, I can boot a Pi to Raspberry OS and arrange for a direct dd from my master Manjaro card to the newer, larger card.
Side note: While performing my first wipe, I noticed dd did NOT automatically expand the partition like I thought it did. Once I have things working better, that may be a possible improvement to look at for a new master Manjaro card.
Prototype GIMP Configuration
First order of business was installing GIMP and XSANE plugin for GIMP. They worked first try, but XSANE only recognized a shared network printer all in one with a scanner built in, ignoring the local USB scanner I cannot seem to arrange access permissions for correctly. I think I spent half my time this week on this problem exploring dead ends.
The most important missing piece to the puzzle has been a script called Divide Scanned Images. With it, I can feed it scans containing multiple pictures, and it can separate and crop them automatically with the option to deskew (rotate to vertical) if an appropriate utility is found. Link to blog about script: how-to-batch-separate-crop-multiple-scanned-photos. Linked on that page in a comment by user Jan is a Linux-compatible version of Deskew (I have yet to get it to work).
Eager to test what I did have working, I went ahead with using the scanner on the network. I had someone put some picture stand-ins on; I got two seed packs. To my annoyance, the separation script appears to only work with pictures on file and not freshly scanned in, making it a completely separate process. As mentioned above, Deskew refused to work. I suspect I either didn’t put it in the right place or I was working with a copy compiled for an x86 processor while on an ARM based system, though it could be as simple as shadows from the seed packs.
Struggling With the Scanner
I find SANE to be an ironic acronym: Scanner Access Now Easy. I still don’t have Linux scanners figured out. I know there’s an easy way with security implications I’ve stumbled my way through before. I also have learned that differences between distros make the Ubuntu page I keep getting thrown to for help useless in parts. Whenever I post for help on Discord, someone else comes along with another question and mine gets buried.
Along my journeys, I’ve learned about a scanner group. I tried adding it to my profile, and somehow managed to get what appears to be a fake group. After a long time trying to figure out how to safely remove it so I could add the real one, I managed to remove myself from all my default groups including my ability to use sudo, and I don’t believe a root account is set up on this card. It even said the incident would be reported. Any such black mark never had a chance to be transmitted over the Internet –WiFi was down for USB scanner diagnostics– before I dd‘ed the master copy back over it.
Another attempt had me searching through the local files for answers. sane-find-scanner and anything looking at the USB port directly can see the scanner right away, but scanimage -L to list devices SANE sees comes up with nothing when off the network. I can’t reproduce my exact path on the laptop I’m working from, but I found a tip to check /etc/sane.d/ for appropriate config files. If I understand epson.conf there correctly, my problem is either elsewhere, or I need both a product ID and a device ID, the later of which I still have no idea how to locate.
Revised Workflow Proposition
In the light of GIMP seemingly not wanting to split pictures live in memory, it may be a good idea to offload that task to a more powerful computer that can handle two Pi’s operating scanners and saving to a network share hosted on an SSD before saving the pictures to GoldenOakLibry. Touch up can then happen in-place.
Takeaway
While I’m glad to have gotten a subpar prototype operational, it’s only about 60% of the process demonstrated at once. Still, it was the missing pieces demonstrated, and the toughest spot was exactly where I expected from past experience. This is already Part 12 of this ongoing series, and I want to finish before another 12 parts go by. Any month may be the month, and no matter how down I feel at the end of work on it, I still made a major goal this week.
Final Question
What was the biggest mistake you’ve ever made but still had a minimal effort fix for?