- Posts: 6
- Thank you received: 1
Did you find a bug? Splash it!
Many false positives in double-page mode
- Copenhagen
- Topic Author
- Offline
- New Member
-
Less
More
1 week 2 days ago #10353
by Copenhagen
Many false positives in double-page mode was created by Copenhagen
In double-page mode, normally YAC tries to detect which images are double-page spreads that should be displayed alone, and which images are single pages that should be displayed in pairs.
I get the impression that in double page mode, YAC reader now often falsely treats many single pages as spreads. It displays such a page alone, centered, when it should be one of two pages displayed. It messes up the layout of subsequent pages and needs to be manually corrected.
This happens even for sources that previously never had this issue, like manga magazines.
I don't really have a repro, beyond:
1. Use YAC in double-page mode.
2. Observe that often only a single page is displayed.
Expected behavior is to display two pages side by side unless the image is definitely a double-page spread.
Note: I use right-to-left reading order if that matters.
The issue is recent, happened after maybe the last update or two. It used to be much more reliable.
Did something about spread detection change in recent versions?
And for that matter, does anybody know what heuristic does YAC use to determine whether an image represents a single page, and when it's a double-page spread?
I get the impression that in double page mode, YAC reader now often falsely treats many single pages as spreads. It displays such a page alone, centered, when it should be one of two pages displayed. It messes up the layout of subsequent pages and needs to be manually corrected.
This happens even for sources that previously never had this issue, like manga magazines.
I don't really have a repro, beyond:
1. Use YAC in double-page mode.
2. Observe that often only a single page is displayed.
Expected behavior is to display two pages side by side unless the image is definitely a double-page spread.
Note: I use right-to-left reading order if that matters.
The issue is recent, happened after maybe the last update or two. It used to be much more reliable.
Did something about spread detection change in recent versions?
And for that matter, does anybody know what heuristic does YAC use to determine whether an image represents a single page, and when it's a double-page spread?
Please Log in or Create an account to join the conversation.
- Luis Ángel
-
- Offline
- Administrator
-
Less
More
- Posts: 3380
- Thank you received: 697
5 days 17 hours ago #10366
by Luis Ángel
Contribute to the project becoming a patron: www.patreon.com/yacreader
You can also donate via Pay-Pal: www.paypal.com/donate?business=5TAMNQCDD...e=Support+YACReader\
Replied by Luis Ángel on topic Many false positives in double-page mode
The single page is not a false positive, is an orphan page before a double page, e.g. [cover] [1,2] [3] [spread], there are two options for 3, repeat 2 and show [2, 3] or show it alone. Users complained about repeated pages when navigating the issue so 3 is shown alone now, and there is a cache that tries to stabilize the layout as much as possible. This obviously depends on the content too.
Contribute to the project becoming a patron: www.patreon.com/yacreader
You can also donate via Pay-Pal: www.paypal.com/donate?business=5TAMNQCDD...e=Support+YACReader\
Please Log in or Create an account to join the conversation.
- Copenhagen
- Topic Author
- Offline
- New Member
-
Less
More
- Posts: 6
- Thank you received: 1
5 days 10 hours ago #10370
by Copenhagen
Replied by Copenhagen on topic Many false positives in double-page mode
Thanks for the reply. You're talking about
this github thread
, yes? I understand the issue and the technical limitations.
But I'm talking about the situation where all images are single-page. Their dimensions are identical. There isn't a single double-page in the entire dataset.
I did some more investigation after submitting the initial post and now I can reliably reproduce the issue. It happens when you scroll so fast that the buffer doesn't have the time to populate, which apparently makes YAC behave as if it's a spread.
So the repro is:
1. Load a large folder in which all the images have identical dimensions.
2. Set YAC to double-page mode.
3. Hold forward arrow for some time.
4. Observe that after a short while, single images start to appear, and subsequent double images are displayed in the reversed orientation. (What should be a left page is a right page, and vice versa.)
Expected behavior is that the layout should remain unchanging in this circumstance. All images except the cover should be displayed in pairs.
If instead you progress slower, by hitting and releasing the forward arrow for each page, the expected behavior happens.
I can supply a dataset if you need it. I used a weekly manga magazine.
Versions prior to 10.4 behaved in a similar way, except instead of single pages, it sometimes displayed duplicates, as you said. So the recent change made the underlying problem more visible.
>Okay so just don't break the reader by scrolling spastically
When you have a 400-page magazine issue but you're only interested in specific 20 pages, fast-scrolling is actually very convenient. One could compare it to flipping through a book.
I am aware that there are difficulties related to file read speed and that preloading the entire folder is impractical. A possible solution I can see is to ensure the next two pages are loaded before making the single/double decision. This is effectively what the user does when he deliberately progresses slower to avoid the issue. So the program could check and prevent going too fast when user is holding the forward button. It doesn't feel like it would gravely affect speed, it would still be faster than manual flipping, and the behavior would be more consistent.
But I'm talking about the situation where all images are single-page. Their dimensions are identical. There isn't a single double-page in the entire dataset.
I did some more investigation after submitting the initial post and now I can reliably reproduce the issue. It happens when you scroll so fast that the buffer doesn't have the time to populate, which apparently makes YAC behave as if it's a spread.
So the repro is:
1. Load a large folder in which all the images have identical dimensions.
2. Set YAC to double-page mode.
3. Hold forward arrow for some time.
4. Observe that after a short while, single images start to appear, and subsequent double images are displayed in the reversed orientation. (What should be a left page is a right page, and vice versa.)
Expected behavior is that the layout should remain unchanging in this circumstance. All images except the cover should be displayed in pairs.
If instead you progress slower, by hitting and releasing the forward arrow for each page, the expected behavior happens.
I can supply a dataset if you need it. I used a weekly manga magazine.
Versions prior to 10.4 behaved in a similar way, except instead of single pages, it sometimes displayed duplicates, as you said. So the recent change made the underlying problem more visible.
>Okay so just don't break the reader by scrolling spastically
When you have a 400-page magazine issue but you're only interested in specific 20 pages, fast-scrolling is actually very convenient. One could compare it to flipping through a book.
I am aware that there are difficulties related to file read speed and that preloading the entire folder is impractical. A possible solution I can see is to ensure the next two pages are loaded before making the single/double decision. This is effectively what the user does when he deliberately progresses slower to avoid the issue. So the program could check and prevent going too fast when user is holding the forward button. It doesn't feel like it would gravely affect speed, it would still be faster than manual flipping, and the behavior would be more consistent.
The following user(s) said Thank You: Luis Ángel
Please Log in or Create an account to join the conversation.
- Luis Ángel
-
- Offline
- Administrator
-
Less
More
- Posts: 3380
- Thank you received: 697
5 days 7 hours ago #10371
by Luis Ángel
Contribute to the project becoming a patron: www.patreon.com/yacreader
You can also donate via Pay-Pal: www.paypal.com/donate?business=5TAMNQCDD...e=Support+YACReader\
Replied by Luis Ángel on topic Many false positives in double-page mode
Ok, yes, I can reproduce the problem, I'll push a fix.
Contribute to the project becoming a patron: www.patreon.com/yacreader
You can also donate via Pay-Pal: www.paypal.com/donate?business=5TAMNQCDD...e=Support+YACReader\
The following user(s) said Thank You: Copenhagen
Please Log in or Create an account to join the conversation.
- Luis Ángel
-
- Offline
- Administrator
-
Less
More
- Posts: 3380
- Thank you received: 697
5 days 3 hours ago #10372
by Luis Ángel
Contribute to the project becoming a patron: www.patreon.com/yacreader
You can also donate via Pay-Pal: www.paypal.com/donate?business=5TAMNQCDD...e=Support+YACReader\
Replied by Luis Ángel on topic Many false positives in double-page mode
Can you give this one a try?
github.com/YACReader/yacreader-dev-build...tag/10.5.1.261111513
Contribute to the project becoming a patron: www.patreon.com/yacreader
You can also donate via Pay-Pal: www.paypal.com/donate?business=5TAMNQCDD...e=Support+YACReader\
Please Log in or Create an account to join the conversation.
- Copenhagen
- Topic Author
- Offline
- New Member
-
Less
More
- Posts: 6
- Thank you received: 1
5 days 2 hours ago #10373
by Copenhagen
Replied by Copenhagen on topic Many false positives in double-page mode
I've run into trouble with installation. I'm on ubuntu.
I've tried running `./build.sh develop` from `yacreader-10.5.1/build_scripts/ubuntu_24.04`, but I'm encountering errors. I also tried `cmake -B build -DCMAKE_BUILD_TYPE=Release`, to no avail. I also tried the .exe under Wine, but it errored out after installation.
I've tried running `./build.sh develop` from `yacreader-10.5.1/build_scripts/ubuntu_24.04`, but I'm encountering errors. I also tried `cmake -B build -DCMAKE_BUILD_TYPE=Release`, to no avail. I also tried the .exe under Wine, but it errored out after installation.
Please Log in or Create an account to join the conversation.
- Luis Ángel
-
- Offline
- Administrator
-
Less
More
- Posts: 3380
- Thank you received: 697
4 days 18 hours ago #10374
by Luis Ángel
Contribute to the project becoming a patron: www.patreon.com/yacreader
You can also donate via Pay-Pal: www.paypal.com/donate?business=5TAMNQCDD...e=Support+YACReader\
Replied by Luis Ángel on topic Many false positives in double-page mode
I'll do a release soon.
Contribute to the project becoming a patron: www.patreon.com/yacreader
You can also donate via Pay-Pal: www.paypal.com/donate?business=5TAMNQCDD...e=Support+YACReader\
Please Log in or Create an account to join the conversation.
- Luis Ángel
-
- Offline
- Administrator
-
Less
More
- Posts: 3380
- Thank you received: 697
2 days 18 hours ago #10379
by Luis Ángel
Contribute to the project becoming a patron: www.patreon.com/yacreader
You can also donate via Pay-Pal: www.paypal.com/donate?business=5TAMNQCDD...e=Support+YACReader\
Replied by Luis Ángel on topic Many false positives in double-page mode
10.5.1 is out, let me know if it’s better now.
Contribute to the project becoming a patron: www.patreon.com/yacreader
You can also donate via Pay-Pal: www.paypal.com/donate?business=5TAMNQCDD...e=Support+YACReader\
The following user(s) said Thank You: Copenhagen
Please Log in or Create an account to join the conversation.
- Copenhagen
- Topic Author
- Offline
- New Member
-
Less
More
- Posts: 6
- Thank you received: 1
1 day 7 hours ago #10380
by Copenhagen
Replied by Copenhagen on topic Many false positives in double-page mode
I tested it and I can no longer get the pages to misalign, so the fix worked. Thank you for the prompt response.
I will however note that truth be told, due to loading, the fast-forward is still not very useful in double page mode. Most of the time the program just doesn't visually update the state at all, so for searching you still need to release the forward key to see where you are; and it can also continue scrolling for a moment after releasing the button. It goes much smoother in single-page mode. But the original problem has been addressed, so treat this as just a note from me. One way or another, I appreciate how responsive you were, and the speedy resolution.
I will however note that truth be told, due to loading, the fast-forward is still not very useful in double page mode. Most of the time the program just doesn't visually update the state at all, so for searching you still need to release the forward key to see where you are; and it can also continue scrolling for a moment after releasing the button. It goes much smoother in single-page mode. But the original problem has been addressed, so treat this as just a note from me. One way or another, I appreciate how responsive you were, and the speedy resolution.
Please Log in or Create an account to join the conversation.
- Luis Ángel
-
- Offline
- Administrator
-
Less
More
- Posts: 3380
- Thank you received: 697
1 day 6 hours ago #10381
by Luis Ángel
Contribute to the project becoming a patron: www.patreon.com/yacreader
You can also donate via Pay-Pal: www.paypal.com/donate?business=5TAMNQCDD...e=Support+YACReader\
Replied by Luis Ángel on topic Many false positives in double-page mode
The flow is there for that purpose.
Contribute to the project becoming a patron: www.patreon.com/yacreader
You can also donate via Pay-Pal: www.paypal.com/donate?business=5TAMNQCDD...e=Support+YACReader\
Please Log in or Create an account to join the conversation.
Moderators: Luis Ángel