tenrec-builders / pi-scan Goto Github PK

My workflow is interrupted with what, I think, is a non ideal frequency of errors during scanning. Here is the log file from a recent scan of approximately 600 pages: https://gist.github.com/meelash/d449ec023bec0bfe9ce5f81861185d52
Errors occur after 45 scans, 80 scans, 10, 27, 16, 25, 74, and 19 scans for an average of one error about every 37 scans. (I'm only getting 433 pages/hour in this run, although it included time taking pictures of the error messages and trying different procedures for clearing errors)

Again, my first question is, are these results typical? Is this something with my setup, or is it a general problem with pi-scan or chdk that I can help to diagnose and fix?

As far as I can see the modes of failure appear to be multiple. There are some times when I get a "Got empty data back when capturing" and I can just hit "ok," rescan, and continue although it seems to have a high likelihood of reoccurring shortly thereafter and requiring more serious measures. Turning the camera in question off and on again seems to always allow me to continue, but sometimes restarting pi-scan, without touching the cameras seems to work. The other evidence I have of multiple modes of failure is the different symptoms on the camera displays. Here are some samples (I can match these up with times in the logs if necessary):

An early error resulted in this memory card error (but I'm not sure if that was there on the initial error, or as a result of restarts and retries to attempt to get past the error without restarting the camera)

Sometimes the error would result in the screen of the camera displaying L3 or L2:


I think in these cases, it would seem like chdk was gone, and pressing the menu button would bring up the normal menu instead of the chdk alt menu:

A restart of pi-scan would return the camera to chdk mode.

There was an error that resulted in the camera becoming completely unresponsive, with the screen black:

Obviously, that required restarting the camera to fix.

Then, there was an error where the camera would make the noise of shutter firing and the display of the camera would show this menu:

So, assuming these are individual issues and not all a result of some shortcoming in my system setup, any ideas which one might be addressable first/easiest? And are these problems related to pi-scan, or chdk?

If related to chdk, then I guess the question is whether they are typical for the performance of other applications of chdkptp or not.

which gpio pins i need in raspberry 3?
how to use them?

i have the original power supply for raspbery
pi scan with battery on cameras seems to work perfect.
i used a power supply with 4,5 volts and 500 mA ,pi scan disconnect the cameras .
if my a3000 canon need 4,3 volts will a power supply with 4,7 volts an 2250 mA will be ok?
cause with 2250 mA and 4,7 volts seems to work perfect
how can i set chdk withour firmware update?
can pi scan be faster?
in any case you have done a great work its more comfortable for me .thanks alot

Now that I have a complete workflow functioning for scanning books and have some experience with it, I would like to increase the efficiency. There are two glaring bottlenecks, currently, one being errors that stop the flow and, the other, general slow performance. I'll be posting another issue to try to isolate the cause of those errors that I get and fix them. Here, I'd like to address the performance problems I'm finding.

Ideally, I think, one would expect the mechanical ability of the operator to be the limiting factor in maximum speed of scanning as opposed to the software. But I find that I am limited by software when I'm scanning.

When I hit the Scan button, almost instantly I hear the cameras fire, and I can begin moving to the next page. But then the majority of the time I can easily complete turning the page and lifting the book back up to the platen well before Pi Scan is ready to capture the next scan. The interesting thing is there seems to be quite a bit of variability in the time spent on the "Capturing..." screen, sometimes it matches my mechanical speed, and sometimes, I am waiting for some seconds.

The performance hit I am talking about is I would estimate that I could easily do upward of 1000 pages per hour physically, but currently I am maybe getting 600-700 even not counting the errors that I alluded to.

I guess the first question I have is, are my experiences typical? Or is there something I am doing wrong?

My system is a raspberry pi2 with touch screen, powershot IXUS 160 cameras, cheap class 4 memory cards (I thought these didn't matter? If necessary I can upgrade them to test).

In spreads, white balance automatically matches whatever the camera is set to, so for custom white balance with the IXUS 160 I just have to set it up manually before starting to use spreads and everything goes as planned.

It seems, with pi-scan, there is a hard-coded white balance that is imposed on the cameras? In spite of having manual white balance on and the correction set using a white card, the scans are coming out with some other white balance. I've attached an example in which the page should be white and the wood in the background should be black. Short-term work arounds would be appreciated if possible.

I have a raspberry Pi model B+ V1.2
Two Canon A2500 cameras
I have a thumb drive plugged into the Pi, plus the 2 cameras and a USB keyboard: nothing else connected.

I downloaded pi-scan-1.0-mouse.img from the link:
Raspberry Pi Image (for mouse)

Wrote the image onto the ssd for the Pi, tried running the software, everything seems fine util it gets to Refocus, I press 3, I hear the cameras take photos, then it gets to generating previews and it locks up, then restarts piscan.

I have tried updating the cards for the cameras: no change, I changed thumb drive: no change.

I am a bit ignorant on how to send you debug files, if you can give me some hints, I'll gladly send them.

I don't know whether this is the right place to ask this but I'm having some trouble with this touch screen (https://www.amazon.co.uk/gp/product/B075JFT9SF) which I want to use with my Raspberry Pi 2 and pi-scan. The touch screen works fine with a normal raspian installation and is calibrated very well however does not work at all with the 'touch screen' version you supplied (I understand this is for the raspberry pi touch screen only) or the mouse version. I have removed the read only permission from the / and /boot mounts and installed the drivers for the screen (https://github.com/goodtft/LCD-show) and the screen resolution is now fine but the touch screen is 90 degrees out - when I press the top right, it shows in the bottom right and when I press on the bottom right it shows in the bottom left etc. I realise this is a very specific issue but could you help at all?

CHDK for many powershot cameras, including the A2500 has a live viewfinder feature. This would make camera calibration much simpler since it the zoom settings could be done in realtime via touchscreen.

the upstream chdkptpgui project supports this with an example GUI written with two Lua frameworks.

The challenge would be to port that code to Python/Kiva

Is it on the forum? In IRC in freenode (#diybookscanner is empty!?)? Matrix, I2p, XMPP!? Ideally would be cool to have some real-time chat public location so as to pick some brains on how to work with pi-scan.

(sorry for issue spam, im awaiting my forum account and want to get this off my chest ;))

Any chance of getting a Pi4 image sometime soon?

tnx.

So, after a while, your pi-scan image gets out of date. By that, I mean, if it is was a typical system, you would run apt update && apt upgrade. However, because this is a read-only file system, we cannot do that. Or is it possible in some fashion? Or is it the idea that you simply re-flash the SD card with the updated pi-scan and get the latest updated OS. Trying to understand the long term maintenance work for pi-scan, can happily submit a documentation patch for this topic.

Hi Jonothan,

I've downloaded the pi-scan 1.0 imaged it to a SD card. But when I boot up it does not recognize my cameras. I'm using Two Canon A2500 cameras that are plugged directly into my Pi 2.

When I click through after the storage option it flashes up that the even camera is not found but then instantly says no cameras found. They are turned on and connected correctly, but the lens never extend out when booted. The SD cards in the camera have been updated but still no luck.

One of the cameras still shows the last picture it took while the other says "no image".

I'm using the keyboard option.

Any ideas?

Thanks
Michael McDonald

Hi, Jonathon. Sorry to bother you again. I'm so close to getting our scanner ready. Pi-Scan is working very well; I got some test pictures on my SD card. Here's the deal: the jpg files don't have rotation info in the exif data. That's my interpretation anyway. Attached are two files, one that is rotated correctly and one that I just took today that is not rotated.

The jpegs have to be rotated to Portrait for my automated crop, OCR utility to work. Any idea if this is a CHDK setting or something to do with the cameras (Canon A810)? BTW, they were working correctly about two years ago when I was using SpreadsPi to capture images.

As a side note, for some reason the picture seems really blueish. You'll see it when contrasting the two pictures. Do you happen to know what that's all about?

Cheers.
Justin

Hey folks, thanks for this wonderful work! I see that the last commit to the project was "22 Mar 2018" and that the #diybookscanner channel on freenode (somewhat related?) is empty, so I am wondering what is the status of the project? I am starting to work with pi-scan and a built scanner and want have a background in software, so I want to see if I can also help co-maintain the pi-scan project to ensure it stays up to date and can progress with the raspberry pi / debian / python etc. moving ecosytems.

Seen from https://github.com/Tenrec-Builders/pi-scan/blob/master/make/README.md, we know that there is a build process to produce a new image but it is not clear how to do that process immediately. I remember messing with debian packaging and chroots before and ending up with getting into difficulty for some $reason on my host computer with mounts. I'd really appreciate some instructions for producing a new pi-scan image.

It would be great to have a CI (like Travis, Gitlab CI, whatever) building the image and also running things like quality assurance checks on the source code (flake8, isort, black, etc.) so as to contribute to the robustness of pi-scan on the dev/software level. Is there any interest in that? I've done this quite a bit and can help. It also makes doing things like experimental features / refactoring easier because you get faster feedback if something breaks in the image building, for example.

Hi there. Downloaded most recent mouse version today (Sept 19, 2017). When booting up the Pi 3 model B I get an error screen that keeps looping.

File "main.py", line 14, in
import wiringpi
ImportError: No module named wiringpi

It says it saved a log to /home/pi/.kivy/logs but I don't see that log file

The same error happens whether I have a keyboard and mouse plugged in or not.

Hi,
made some attempt to use pi-scan with 2 of Sony 7rII.
They are not recognized at startup but, using the console (after pressing 9 several time) and then gphoto2 we were able to take pictures, get the cameras parameters and download images. Is there any special config to allow the pi-scan software to recognize those cameras?
Many thanks
Giuliano

Hi there,

PiScan is loading, finding my SD card, and finding the two cameras. When it gets to the screen where you can adjust the zoom, focus, etc., whatever option I want to do ends up with the same error and failure to proceed. This is what displays on the screen twice, once for odd and once for even cameras.

Failed: Failed to prepare camera: Failed to disable flash

I'm not sure which log file to give to you to help troubleshoot this. I'm wondering if I've installed CHDK properly, though each camera boots up just fine into CHDK. Both cameras are the same model, Canon PowerShot A810

Thanks for your help.

AFAICT, pi-scan still works on Python 2 and the raspbian image it builds too. Python 2 is EOL'd however and raspbian ecosystem is moving along. The longer pi-scan leaves upgrade work undone, the potentially more difficult it can become later on.

So, I am wondering if there is some motivation to migrate the code base to Python 3. Typically this migration process is low priority but I thought I'd raise the issue regardless.

#17 is related (if I could build new images, I could start to experiment with this migration process).

Should be very simple to add a button to move the existing images into a folder named, "book 1", "book 2" etc. thereby emptying the images folder and resetting the pages count so it can match the book currently being scanned when running through multiple books in one sitting.

I'm quite close to getting pi-scan running on OS X in which case it will be easy for me to contribute to it and have a good debugging and testing environment, in which case I will do this and submit a pull request.

A wish list item is for voice command whereby you can speak a button label and pi-scan executes the action. This is most helpful for "Capture", but would be great for all buttons.

Steps to duplicate:

1. Connect storage and cameras and power-on
2. Begin Scanning->Next
3. Debug Cameras(4)
   Display shows:
   Camera Crashed. Turn camera(s) back on and click ok
   and Even and Odd Camera are both Searching...