"Let’s say you’re writing an image processing program. The program takes in an image, converts it to floating point, does some processing and finally saves the modified pixels to disk as 8-bit colors. "
excuse to argue about the best way aside, if this is the goal you should not be rolling your own image file reading. you should use openimageio. idk what approach it takes in its internal conversion to float, but that library is more likely to have the right answer than you trying to roll it yourself given its the library used internally by tons of professional image manipulation software...
If you're a beginner, or just want something which works quickly, sure.
However OIIO is far from perfect in all situations (having had to debug and fix issues with its mip-map generation filtering code in the past), so don't always assume that just because there's a mature open source library out there doing something that it's always perfect.
sure of course nothing is perfect and oiio has a lot of surface area / is still oss. thats good advice.
ive just seen a lot of "ai researchers" who are getting into professional image processing and are both beginners and want things quickly and so could do much worse than just starting from what they get out of oiio. especially for a lot of the non-obvious stuff (more of that in color handling than just the io stuff though)
pull up any moderately busy picture with more than a trivial amount of objects. pictures of "traffic" or
with other similar repetition are great for this demo. pick one specific object (like a specific tire on one car) in the image and write (or say) out all the words youd need to specify that exact object. now take the same image and point at the object with your mouse or circle it with an annotation tool. its often very very hard to describe accurately which object you are talking about, you will often resort to vague "location" words anyway like "on the upper left" that try to define the position in a corse way that requires careful parsing to understand. pointing/annotating is massively superior both in brevity, clarity, and speed.
Not sure. I never had to work with Red footage to that extent. I just remember the colorist showing it to me. Maybe he knew about it and just pointed it out to me because it was such an extreme example. I wasn't a colorist, just someone that worked with a colorist. An assistant would even be glorifying it.
what dictionary does this use, i tried a few words that i was pretty sure are in the scrabble set but didnt work here. (could be wrong)
for me the train is a little bit too fast for the letter input method. like even if i have a word by the time i get all the tiles down the train has made more progress than my word added. some kind of way to get a bonus for longer words that gives you more time feels like it would be a nice balance tweak. like risky, take more time to find a longer word but more payoff for doing it. basically makes me look at the rack and say... do i have a long word here... do i have time to find it...
also let me rearrange the tiles on the board not just my rack. its very frustrating if you need to swap two letters you have to move the first one out of the way then put the second then move the first again.
if op is really serious about fixing social networking he needs to figure out a way to operate it that wont enshittify.
ie, public good, not for profit, something like that.
as is hes already signaled he intends to enshittify it eventually ("premium features"..?) which to me is a non-starter.
the problem is that a successful network beyond a certain size like this needs funding. its unfortunate but this needs legal /compliance, moderation, even marketing...
those things aren't free, but you could imagine ways to pay for them. Id totally accept a small subscription fee for a network like this if paying that fee could guarantee privacy, and that the person hosting all my data would not be looking to squeeze every dime of value out of me as a user.
anything less than solving the funding problem and you are just saying you will become facebook (or get bought by them). no thanks.
the full on duino mod is way better for one primary reason. power steam or whatever they call the thing they do with the steamer. it takes the sputtery slow and weak steam wand and makes it actually functional.
also the "adaptive" settings you can do on there let you basically never have to "dial in" a shot. throw any random (still decent q and paired with a good grinder) beans on there and you get a shot thats very drinkable. For lattes and stuff where you are hiding the coffee in milk anyway this is like still better than 90% of what you can get from the coffee shop with the 20k$ big iron and 0 effort. and when you do have that special bag that wants the extra attention and straight espresso you've got all the controls and presets and shot tracking.
if you are gonna try and do home espresso you already have an involved hobby, if you are gonna mod your machine you are already pretty hardcore about your hobby... you might as well go all in at that point
the most interesting thing here is actually maybe that this is cad targeted for easy 3dp modeling. while you wont ever unseat the big guys there is a lot more you could do to make this sub space better by targeting it.
like could you make it aware of the fdm limits and help me avoid them while im building instead of needing the build-slice-refine flow across multiple software?
Yes - supporting full manufacturing intelligence is part of the larger vision (in practice quite far off).
I don't think there is need to unseat anyone. 3D modeling market is expanding and the intent is to serve people for who are not users of current market leading tools. There are tons of plausible UX paradigms that have not been explored. This is one such exploration :)
"like could you make it aware of the fdm limits and help me avoid them while im building "
That's part of the long term vision. First solve modeling, then solve manufacturing of the models reliably.
The modeling already follows this principle - you can't model things the rest of the operations can't support.
Yes and it was a massive manual effort. In a way they acknowledged that keying does not really work all the way and having that unnatural color everywhere in the set is not worth it. It’s a massive production with heavy VFX work so not something you can apply to your own production. Sand screen and roto sections of this discussion are interesting.
this is the approach that stop motion uses, except they get to keep the camera in the same place. its still not perfect because of spill from the background onto the foreground and requires additional masking and cleanup.
im familiar with this work and specifically they tried replicating the sodium vapor style approach but what worked for poppins level isnt actually good enough for today. Specifically you still end up with light spill that contaminates the foreground, especially for things like the fresnel reflections on the side of a face. the magenta idea was to still do what is basically a color difference key, but increase the color separation between fg and bg by lighting the two with different opposite colored lights. then using a ml model to recover the original fg color.
excuse to argue about the best way aside, if this is the goal you should not be rolling your own image file reading. you should use openimageio. idk what approach it takes in its internal conversion to float, but that library is more likely to have the right answer than you trying to roll it yourself given its the library used internally by tons of professional image manipulation software...
reply