That's a really nice write up! I'm sure it will help a lot of people. Below I've given some feedback on some of the points you raised that might help others.
GenioDiabolico wrote: ↑
26 Sep 2018, 10:00
Using the Siconi mat, I don't adjust the book at all at any point. I push it down as hard as I can at the beginning, and it holds all the way through.
This is good news. I think I'll have to look into it. After Daniel mentioned something similar way back (here: viewtopic.php?f=14&t=3007
) I looked into them but never got around to testing them.
I also found this product (that uses "nano-suction"), but it seems they only sell it in bulk and I hadn't found a retail supplier for it then (perhaps there is one now). They say 6" square will hold 10lbs vertically:
http://www.everstik.com/everstik-nanosu ... works.html
GenioDiabolico wrote: ↑
26 Sep 2018, 10:00
I am standing on the same side as the arm but that is not generally a problem. I'd rather have it like you do but the right side of my desk has no lip to clamp it to.
Have you tried the back of your desk (as you'll see where I have it for when I'm not using the lamp). You'll may need to add the extension pipe to push the camera forward so it's nearer you and you're not having to lean over the table to get the platen under it.
GenioDiabolico wrote: ↑
26 Sep 2018, 10:00
Looking at the timestamps of my most recent run, I did right side batch of 204 pages in 18 minutes, the left side also in 18. That gets me to ~680 pages / hour but that is all in from beginning to end of the main scan run. A few files just didn't take so there were a few minutes of figuring out the missing ones and filling in. If you are adjusting the book every 30 pages, what is your actual realized throughput from beginning of book to end?
I posted the time stamps from my first real test with this new design (i.e. using the smartphone and Bluetooth trigger) as compared to the original version using a point and shoot (with no focus lock). I did 122 right hand pages of the book in 8.2 mins.
Mohib wrote: ↑
26 Mar 2017, 03:56
Just did my first real scan test of a hard cover book (right hand pages):
122 pages in 8.2 mins (start to finish, all in) (time stamps attached)
= 14.9 pages/min
= 892 pages/hour
That 8.2 mins includes twice pausing scanning to adjust the macro focus rail a few mm down (per the reasons given in my original post for my scanner).
Today I'm faster than that because:
a) I've had more practice and am able to scan without moving the book as much (and, from what you've said, the Siconi mat now all but eliminates the book shifting), and
b) I don't need to adjust the macro focus rail (which I used to adjust to try and keep the page image scale consistent but now I normalize the image scale with the software scripts). Also there was time wasted refocusing after adjusting the macro focus rail.
I think b) accounted for about 45-60 seconds of the 8.2 mins, so say 45 seconds to be conservative, that would put it upto around 985 pages/hour on my first test.
I think the Scioni mat will probably cut off another 30 seconds of fiddling with the book and take it upto about 1050 pages/hour or about 17.5 pages/min.
So that was back then on my first real test.
One thing I think that makes a big difference is your own state of mind and the scanning environment. Given the human element involved, since you are "part of the scanner" machine, a clean, clutter free, obstruction free desk with ample elbow room that allows for fluid page turning without any distractions that you have to focus on makes a big difference. I think having the scanner post in the front, rather at the back, would slow me down quite a bit. You might want to try moving it to the back of the desk or trying another desk or dining table that allows it to be clamped to the side.
Practice since then has let me learn how to "pre-click" the blue-tooth trigger. So I actually press the button while I'm still bringing the platen down so the slight delay between when I click and when the camera shutter fires is eliminated. If you look closely in the hard back test, in my video, you'll might just be able to notice me doing this "pre-clicking" in the last 5 pages.
You mentioned similar a delay in your post (along with time lost from autofocus):
Configuring your camera app so the camera is snapped as quickly as possible helps. Mine takes too long and adding a second per photo adds at least 5 minutes to a 300 page book. I’ve tried different autofocus options to speed this up. Thus far I have had limited success.
Manual focus or focus lock is essential in cutting down camera delays (and was the big delay I had with my point and shoot in my original design).
Pretty much all scanners that have a max rating of x pages per min will lose about 25% of that max speed when scanning a full book due to all kinds of things that pop up (pages sticking, fatigue, etc.) So I think 1,100-1,200 pages per hour (18-20 pages per min) is probably the best I think will be possible with the TIFLIC scanner (assuming hardcover books).
I've timed videos of many twin camera setups I've seen here on the forums, and they manage about ~1,000 to 1,400 pages per hour (IIRC Daniel's scanner was clocking in at about 1,600-1,700) because a lot of time is wasted between turning pages (because the motion to lift and drop the platen is quite large and time consuming). Note also, these would be around their top speeds as timed from a brief 1-2 min scanning session shown in the videos and not the actual net speed from doing a full book with any complications that may arise (stuck pages, etc.).
So given TIFLIC's simple construction and single camera operation, I think the top speed at around 1,500 pages per hour and a net speed of 1,100 to 1,200 pages per hour is an excellent result.
You also mentioned in your post:
I have tried to get it so that ScanTailor can do all its work on automatic. My first few attempts when it was drifting a lot were labor intensive, now I just let ScanTailor place the boxes and they almost always work. The end results are quite positive and getting better.
That's good to hear. Yes, ScanTailor is very good and the better the original images the better job it does.
Now, more from you post ...
... pick 6 holes if you want to buy exactly what is in the plans. It would be possible to in fact do 4 or 2 holes, depending on the flexibility you desire from your platen. In practice, I’ve been using mine for two weeks and have yet to move the handle to different holes. It costs $2 per hole drilled, so I gambled $8 on future flexibility.
I rarely move my platen handle also. I usually have it attached to the holes closest to the centre (i.e. attached to hole #1) because that lets me provide the most even pressure on the page and also have enough leverage to press the page flat at the spine too. The main reason for the extra holes was to get rid of handle reflections (if you're not processing with Scan Tailor) as I explained here:
All of this will be presented below. Mohib’s plans are sufficient to build a scanner but what they lack is an Instructable style “Step 1 – do this” flow. I will attempt to put a little of that in.
Yes I know that's missing from the instructions. I was planning on writing them up, but never finished it. However, I do have quite detailed, step by step, instructions in my original version, though they need to be adapted to the new version as there's less to do. You can get the original instructions here:
This is where those times spending 10 minutes finding me the right 17 cent screw got repaid, as I ultimately dropped around $100 in materials and tools there working on the scanner.
I hear you! I spent hours at the local hardware stores looking for various nuts and bolts, and going through lots of iterations till I finally got the right ones!
One of the parts in question was the barrel nut. This is those little things you find in Ikea style flat pack furniture, cylinders with a slot for screwdrivers on the ends and a threaded hole through the center. The plans call for two with 1/4-20 threading and a length of 1 1/2", which are hard to come by in the US. The plans cite a supplier in the UK but shipping is prohibitive for these things. I wanted to avoid that. Luckily, True Value had the same thing but in 3/4" length. I decided to experiment with this and only buy the UK version if absolutely necessary. Spoiler Alert – the 3/4" version worked.
Yes the barrel nut is a problem to get. I got mine from Ikea after a lot of begging at their spare parts dept!
Glad to hear the 3/4" version worked because I never tried it. However, I can see how it would make getting the steel cable attached a problem since the 3/4" version can't anchor into holes in both sides of the 45 degree elbow. That's sort of essential for the handle for the cable to stay in place when the handle is not attached to the platen itself.
... I won’t lie to you, the platen handle is by far the most challenging part of this whole thing. The steel cable holding it together is the worst bit of that so once you make it past that, it’s all downhill.
Assuming you have the 1.5" version, I had posted this "assembly video" of the platen to show how I get the cable in place:
I also went through the plans and ordered most of the things from Amazon, such as the tripod ball head, the focus rail and the star knobs. Where I think Mohib is crazy is in not using affiliate links to Amazon because he’s just leaving 4% of all those purchases on the table. I’ll list out all the stuff here with my affiliate link because I am not crazy. If Mohib ever lets me know of an associate ID for him, then I will swap them out for his because he deserves the money more than me. However, someone does deserve it.
That's kind of you, and once I've figured out how to get setup and have my affiliate account I'll let you know. 4% will pay for coffee!
Step 3 – Drill a 1/4” hole in the center of the 1 1/2” PVC caps. Threaded rod will go through these. From the photos in the diagrams, it looks like the cap that connects to the clamp has extra holes and extra bolts in it. I did not drill these holes and had a few bolts left over from the original manifest. I am not sure what to do with these, so I ended up skipping.
I had some incorrectly drilled holes in them, but there is really just one extra hole needed in the cap that fastens to the Manfrotto clamp (see page 29 in the PDF instructions). That extra hole is for that small M5x20 (20mm length) and goes into the Manfrotto clamp to prevent the pipe cap from turning because if it does, I find it ends up unthreading the rod and the whole thing gets loose.
A note on the color of the materials: careful readers will note I mentioned getting white PVC parts for the camera arm while noting those parts are black in the photo of my camera arm. I found early on that the white pieces showed up in reflection on the platen, so I spray painted them black.
... Mohib lists his LED lamp as optional in the original plans but I found in my early tests that I wasn’t getting enough light on the book and it was very subject to shadows.
Ok now with respect to reflection, lights, focus and OCR, I did extensive tests (see posts below (and more in the threads after the ones below) for results; some are with my original scanner and point and shoot camera).
In general I found that if you're post processing with ScanTailor, as long as the light is "reasonable" and you're not getting a real "mirror like" reflection, they all disappear. As you can see from the images in the posts below, I'm not scanning in a darkened room, using daylight or general incandescent room lighting (without worrying about white balance), have lots of reflections and even the camera shows up in the image, but it all drops away during Scan Tailor processing without any impact on OCR.