[gwern]

Have you considered, since you can extract the images via the mask, selectively inverting them?

One can fairly reliably use a small NN to classify images by whether they should be inverted or just dimmed, and I've used it with great success for years now on my site: https://invertornot.com/ https://gwern.net/invertornot

---

On a side note, it'd be nice to have an API or something to let one 'compile' a PDF to dark-mode version PDF. Ephemeral browser-based is a drawback as often as a benefit.

[simoneamico]

That's actually exactly where I started. The initial idea involved a YOLO nano model to classify images, deciding what to invert and what not to. It worked as a concept, but during the feasibility analysis I realized that for native PDFs it wasn't necessary: the format already tells you where the images are. I walk the page's operator list via getOperatorList() (PDF.js public API, no fork) and reconstruct the CTM stack, that is the save, restore and transform operations, until I hit a paintImageXObject. The current transformation matrix gives me the exact bounds. I copy those pixels from a clean render onto an overlay canvas with no filters, and the images stay intact. It's just arithmetic on transformation matrices, on a typical page it takes a few milliseconds.

Your approach with a classifier makes a lot more sense for the generic web, where you're dealing with arbitrary <img> tags with no structural metadata, and there you have no choice but to look at what's inside. PDFs are a more favorable problem.

A case where a classifier like yours would be an interesting complement is purely vector diagrams, drawn with PDF path operators, not raster images. Veil inverts those along with the text because from the format's perspective they're indistinguishable. In practice they're rare enough that the per-page toggle handles them, but it's the honest limitation of the approach.

[gwern]

> In practice they're rare enough that the per-page toggle handles them, but it's the honest limitation of the approach.

I don't understand how you handle raster images. You simply cannot invert them blindly. So it sounds like you just bite the bullet of never inverting raster images, and accepting that you false-positive some vector-based diagrams? I don't see how that can justify your conclusion "it wasn't necessary". It sounds necessary to me.

[simoneamico]

Actually, raster images are never inverted, they're protected. The CSS filter: invert() hits the entire canvas (text and images together), then the overlay paints the original image pixels back on top, restoring them. The result is: inverted text, images with their original colors.

The choice to never invert raster images isn't a compromise, it's the design decision. The problem veil solves is exactly that: every dark mode reader today inverts everything, and the result on photos, histology, color charts, scans is unusable. Preserving all images is the conservative choice, and for my target (people reading scientific papers, medical reports, technical manuals) it's the right one.

It's absolutely true that there's a subset of raster images, like diagrams with white backgrounds and black lines, that would benefit from inversion. I could be wrong, but in my experience they're a minority, and the cost of accidentally inverting the wrong one (a medical photo, a color chart) is much higher than the benefit of inverting a black and white diagram, from my point of view. For now the per-page toggle covers those cases.

[gwern]

> It's absolutely true that there's a subset of raster images, like diagrams with white backgrounds and black lines, that would benefit from inversion. I could be wrong, but in my experience they're a minority, and the cost of accidentally inverting the wrong one (a medical photo, a color chart) is much higher than the benefit of inverting a black and white diagram, from my point of view. For now the per-page toggle covers those cases.

OK, so I did understand, but this sounds very hand wavy to me. You say it's a 'minority'; well sure, I never claimed that was >50% of images, so I suppose yes, that's technically true. And it is also true that a false positive on inverting is usually nastier than a false negative, which is why everyone defaults to dimming rather than inverting.

But you don't sound like you have evaluated it very seriously, and at least on my part, when I browse my dark-mode Gwern.net pages, I see lots of images and diagrams which benefit from inverting and where I'm glad we have InvertOrNot.com to rely on (and it's rarely wrong).

It may be nice to be able to advertise "No AI" at the top of the page, but I don't understand why you are so committed to biting this bullet and settling for leaving images badly handled when there is such a simple easy-to-use solution you can outsource to, and there's not a whole lot else a 'dark mode PDF' can do if 'handle images correctly' is now out of scope as acceptable collateral damage and 'meh, the user can just solve it every time they read every affected page by pushing a button'. (If Veil doesn't exist to save the user effort and bad-looking PDFs, why does it exist?)

[simoneamico]

It's not resistance toward AI. Machine learning isn't among my current skills and I preferred to build with tools I could maintain and debug on my own, but the door isn't closed. Thank you for pushing on this point.

[AbanoubRodolf]

The raster image problem is real but there's a middle ground between "never invert" and a full NN classifier.

The author already computes BT.601 brightness per page. You can run the same calculation per-image bounding box instead of per-page, then add a bimodal pixel distribution check: if a raster image has most pixels near black or near white with few midtones, it's probably a line diagram or screenshot, not a photograph. That heuristic catches the main false-positive case (black-line diagrams on white backgrounds) with maybe 20 lines of image processing code.

It won't be perfect, and gwern's point stands that a proper trained classifier would be more accurate. But for a PDF viewer where you're already parsing content streams to get image coordinates, it's a lot cheaper than shipping a model and handles 80% of the problematic cases. The remaining edge cases (medical scans, thermal images) are rare enough that the per-page toggle is reasonable fallback.

[gwern]

FWIW, we did consider a histogram heuristic, and I believe GreaterWrong still uses one rather than InvertOrNot.com. But I regularly saw images on GW where the heuristic got it wrong but ION got it right, so the accuracy gap was meaningful; and that's why we went for ION rather than port over the histogram heuristic.

[simoneamico]

Really appreciate this AbanoubRodolf, thank you. The brightness analysis code and the image bounds are both already in the project, I just never connected the two. The distance between where I am and where you're suggesting I go is really short. Feedback like this is exactly why I posted here. Thanks again