Modern Image Processing Algorithms Implementation in C | Hacker News Reader

Modern Image Processing Algorithms Implementation in C

sod.pixlab.io

199 pointssymisc_devel25 comments

Since it seems like the poster works for the company: the ads on this article are all semi-nsfw for me, more-so than the average GAd I see (is that something you can tune?). I reported them to Google but that might be a black hole.

Additionally, why in the world does a technical blog for a for-profit company need ads?

I got ads for LEGOs because I have kids and have recently searched for LEGOs [this is less of a humblebrag and more of a sigh at how old and boring I've gotten.]

I'm curious if the degree of ad personalization varies by site (do some sites show more personalized ads?). Or maybe the average demographic of a site's viewers is somehow factored in?

The site uses in the code the javascript for Google Ads and GoogleAnalytics, so probably you searched for LEGOs in Google search. Or was in other search engine using Google Chrome?

The Google Ads algorithm is not exactly known, there are guessings, but the so called "auction" that made you to see it probably is not page content related.

Hello,

It's a blog post by one of our engineers. Authors are free to monetize their content without inference from the company.

That sounds like a smart way to get your engineers to take the time for writing! I do assume you all do at least a basic suitability and security review, though?

Of course. Only Adsense is authorized. After a short investigation, it appears that the main reason these NSFW-limit ads are shown is because the article includes direct link to the PixLab NSFW API Endpoint (https://pixlab.io/cmd?id=nsfw) which is basically a bridge to our ML model that let you detect whether an image or video frame contains adult, bloody or gore content.

Sorry if that came across like I was questioning that part of your process, I was more curious in the actual process, but it is neat that you guys already managed to figure that out!

Getting shitty ads is a by-product of privacy consciousness.

When I stopped clearing my cookies periodically, the ad quality went WAY up.

Also, I'm guessing that image processing is moderately related to porn.

Same here

I got three images, two of pretty college women, and a third one that I can’t tell exactly what it is, but it’s skin color and my first impression was very NSFW

It looks promising.

I like the way each algorithm is explained with minimal example and demo images.

I'm looking for such a lightweight vision lib to embed simple image manipulation programs on tiny ESP32-CAM boards.

OpenCV seems too heavy to integrate on such small devices.

So far, I've been able to develop simple image processing programs with CImg[1] (simple filtering, image cropping, adding text).

I'll try to spend some time exploring this SOD lib, thank you for sharing.

[1] https://cimg.eu/

Visit the url below for much of the same content but without ads. https://sod.pixlab.io/

Note this is GPL which will make it a harder sell vs OpenCV in many cases. I'd guess they're looking to use the non-GPL for $$ model.

Nevertheless, it looks interesting. It looks really nice for embedded where you don't always have the GPU for compute.

Is there anything substantially different here from the way any modern post-processing shader works right now? For example: https://threejs.org/examples/webgl_postprocessing_rgb_halfto...

Don't know about the algorithms, but this is a C library: for computers. threejs is a javascript library: for browsers

How does it compare to dlib [1], ccv [2] etc.?

[1] https://github.com/davisking/dlib

[2] https://github.com/liuliu/ccv

I like dlib better if you want a swiss army knife of algos.

ccv is moved on to be all-in on neural-network based computer vision and all development effort happened in that area. (Similar can be said for dlib but at least the author still implement random black-box optimizers from time-to-time).

SOD seems try to reimplement some of OpenCV but TBH, these are not "modern" at all. These are 15 year-old algos while works, the useful domain is shrinking and people who interested in these also might be looking for hardware-accelerated options (AVX-512 / CUDA etc). Some very specific classic heavy CV applications (such as SLAM) are quickly moving to their own dedicated libs (such as Cartographer from Google or ORB-SLAM).

Anything like this but for photo imaging manipulation?

Thanks for sharing!

better than OpenCV ?

The focus seems to make a smaller (not better) OpenCV

easier to integrate into C-based stuff, probably...though unfortunately doesn't seem to make it easy to work without dynamic allocation, which makes it not as useful as it could be...

Curious to know as well

Better in what sense?

I mean OpenCV is C library, lots of well tested algorithms, Python binding, lots of users, docs, examples, you name it

To be considered as potential use, it should either do something which OpenCV is not doing, or do things (a lot) better.

I mean, I used OpenCV, shall I move to that lib? OpenCV is pretty much standard C imaging library, anyone working with images knowns about OpenCV

OpenCV is C++ (though there are 3rd party C bindings)

Reply on news.ycombinator.com