Image Manipulation (

Publié le: 06 octobre 2022
sur la chaîne: KrossD Line

CHAPTERS:
Episode Intro ~ 3mins
Image processing in Javascript ~ 4mins
Install "Pillow" for processing in Python ~ 33mins
Image processing in Python ~ 37mins

In today's machine learning diatribe I cover image manipulation. This project is aiming at relying heavily on graphics engines as stated in earlier videos. Creating and handling lots of images is therefore going to be crucial. Since I'll be using both javascript and python for my programming languages of choice throughout the course of this project, I figured it'd be best to introduce the basics of both programming languages so that I can readily prototype in one language and implement in another language in later videos without needing to stop to explain jumping between them at length mid-video.

I start off with javascript since that's what I'll be using for most of the concepts and prototypes of ideas I want to try. I'm using HTML5 canvas, as always, for simplicity's sake and to minimize the issues surrounding versions or library vulnerabilities down the line. This will make these code examples very accessible since almost any browser should be able to run the same code as opposed to other platforms which are sometimes complicated by IDE's, environment setups, installation procedures, version controls, etc. By running concepts quickly in the browser with no plugins required, and with the sandbox'ed environment built in to the browser, I figure doing as much as feasibly possible in a web browser first will ultimately make my code and project notes more available to others who want to try some of these snippets out for themselves.

In this video I also spend a couple of minutes updating Anaconda python environment and installing Pillow, the library I use for image processing in Jupyter Notebooks. I'm sure there are many other ways to perform the same tasks, including other libraries or even manually running through streams / bitmaps with just numpy by itself, but I think this is the most effective way without introducing a load of overhead or dizzying amounts of code to accomplish simple tasks. Performance seems pretty decent so far and the library is pretty intuitive for the most part, despite appearances and my stumbling about.

After the javascript demo and the python installation of the Pillow library I play a bit with automating some image processing in python. For this example I read and write bytecode image buffer data just like in javascript earlier in the video, but I take it a bit further and demostrate some batch image resizing and quality adjustmnets. I forgot the filtering demo I wanted to do, but essentially the same process of iterating through files in the current active directory but just using filter(GaussianBlur(blurRadius).save(filename) instead of just .save(filename). The pillow documentation is excellent and I'll leave a few non-affiliated tutorial links below that I found particularly helpful). This is an excellent little script even for people who aren't interested in code or machine learning research, I should note, since most people have more pictures on their phones than they'd care to rename manually by hand. I suspect over the years people will accumulate pictures that will only get harder and harder to organize and manage. Having an idea of how to run a script to processing a huge batch of images even if only for resizing or changing the filenames or something is an incredibly valuable intellectual tool to have. When it's needed it could save tens of hours of tedium at some point in your life, or save you from risking identity theft if you one day try downloading that free software to automate this sort of task for you.

I should also note that my chronic health issues have a tendency to affect my speech somewhat, so I apologize for the brain farts and slower speaking. I recorded this at 7am at my most lucid so this is pretty much as good as it gets until late Spring so hopefully it's not too jarring for viewers. If it is a distraction though, do let me know in the comments and I'll maybe consider releasing videos less frequently but I'll script them a bit ahead of time to help me keep my tongue moving.

If you'd like to support this kind of content and keep my research free and ad-free just subscribe and stay tuned. No hocking of wares, no promises of freebies on an exclusive discord or anything like that. We need more clean content on youTube, I'm making a very public statement with this project.

My Links:
This research project & channel's twitter account:
@KrossDLine @paxdriver
www.krisdriver.com

Non-Affiliated Episode Links (not affiliated, I just found them helpful):
https://www.adamsmith.haus/python/ans...
https://www.geeksforgeeks.org/python-...

Intro music produced by Kid Flash
Setup up for python interlude - "Produced by Don P"
Outro music produced by Don P

Sur cette page du site, vous pouvez voir la vidéo en ligne Image Manipulation ( durée heure minute seconde en bonne qualité , qui a été Téléchargé par l'utilisateur KrossD Line 06 octobre 2022, Partagez le lien avec vos amis et connaissances, sur youtube cette vidéo a déjà été regardée 37 fois et il a aimé like téléspectateurs. Bon visionnage!

5,147

474