I mean, I did originally nominate it for the Intermediate Book Club, but people thought it’d be better over here.
4 Likes
I never vote for them, but every time you share an interesting fact about your nomination, it pulls me in, and I end up joining the book club! 
To all members: Please don’t vote for brutal-level hard 4-koma anymore. Didn’t you learn your lesson last time?
7 Likes
I made a home thread.
15 Likes
I’ll see what I can get together.
What about brutal non-4koma?
12 Likes
By the way, do you have a single place with all the stats you’ve gathered on various manga?
4 Likes
I believe so 
5 Likes
Ah, right. Looks like there’s no way to filter by whether a manga is 4-koma or not.
3 Likes
Sounds almost like a feature request 
5 Likes
It’s on the to-do list.
The difficult part about tagging as 4koma is that I don’t know an easy way to get this data other than looking through the manga to see which ones are 4koma.
8 Likes
Use OCR to look for consistently square or rectangular panels on each page. That’s assuming you’ve been doing some custom OCR stuff and not just using a library out of the box. I bet you could train a neural net to determine if a manga is a 4-koma. 
3 Likes
Wouldn’t a simple flood-fill of the white portion outside the rectangles work a lot of the time? Then you just count the number of areas you’re left with.
4 Likes
So I think it would work with 紡ぐ and Bocchi, now whether it could work generally is a different story… I’d need to test on a much larger sample size.
This code is very slow but it’s a proof of concept.
As you can see in the examples below of course it messes up a bit as soon as there’s stuff drawn between the cells (which is not uncommon in these 4koma) but I think if you sample a bunch of pages and you have a well tuned heuristic it could work (something like maybe “you have at least 5 cells with almost exactly the same size then it’s probably 4koma”). You could also check that the regions line up in a grid for extra safety I suppose.
examples
This one works fine, one of the text bubbles extends outside of the cell but there’s no overlap:
This one has a text bubble overlapping, so my naive fill algo merges both:
Another overlap, but this time it works out because there’s enough of a delimitation:
Random Dungeon Meshi example for an example of a negative result:
5 Likes
Actually an easy way to “optimize” the computation is simply to aggressively downscale the image before running the detection. It may even help remove noise and inaccuracies.
1 Like
This is the route I ultimately went with (after improving my existing code based on OpenCV).
I went the route of “if there are eight panels, four on the left half, four on the right half, it’s a 4koma”.
There are also 4koma with four wide panels per page, but I haven’t tried handling that yet. Should be easy to do. (デキる猫は今日も憂鬱 is an example of this.)
But for now, at least I have a start. (And for anyone using the site to track known words and see their stats for manga, there are pages for IMC vote winners and IMC active nominations which I use to gauge how painful each entry will be for me.)
9 Likes
TBH, a wall of text isn’t even that hard if you’ve ever read a novel. But indecipherable handwriting, uncommon words, all kana, and dialects that even the characters in the same manga can’t understand… that’s another story! 
10 Likes
That’s true, but I also think that this IMC is used as a stepping stone before reading novels (that’s what I did, at any rate), so you can’t assume that everybody here is comfortable reading novels.
Also, as you mention, you have all the manga-isms of dialectal speech, very casual Japanese, confusing dialogue where you’re not sure who’s talking etc…
That being said I think we’re probably overdoing it a bit in this thread with the big WARNING signs, I think it’s good to manage expectations but at this point we’re probably actively discouraging people from participating.
It’s hard but it’s not that hard.
12 Likes
Oh no, it’s not my intention to discourage anyone from reading. I might complain a lot, but as long as you enjoy a work, you can overcome any challenge. If you find novels overwhelming, try starting with a genre you love. For me, mystery works best. And as for 4-koma, a volume of Nozaki-kun has around 35k characters, and we read about 7k a week—but no one ever wanted to slow down the pace. They don’t want to take a break between the volumes these days. That’s how much we love it!
I believe there’s always a work that’s worth all the effort.
8 Likes
Also, I need more of you to join the dark side and read 4-koma nonstop.
6 Likes
魔入りました!入間くん
Natively: https://learnnatively.com/book/a4870183c8/
Summary
Hopeless pushover Iruma Suzuki has found himself in a devil of a predicament… His trashy parents have sold off his soul, and he now has to live and attend school in the Netherworld. But with his unique survival skills and doting demon grandfather’s support, Iruma will surely make it through this hellish experience. He’ll just need to subjugate rival classmates, summon familiars, and more, all while never revealing that he’s human… Easy as aleph, bet, gimel, right?
Availability
Amazon JP | CD Japan
Kindle | Kobo | BookWalker
Personal Opinion
I’ve read this, and thought it was a really fun comedy. There’s also an anime for those interested in that. I really like protagonists who are too nice for this world yet somehow still make it through perilous situations, and Iruma fits that description to a T.
Pros and Cons
Pros
- The various demons have pretty unique speaking styles, so you’ll get a dabble of practice in a lot of things.
- Lots more reading material if the book really speaks to you.
Cons
- The comedy may not work for everyone.
Pictures
19 Likes