Here are some photos from last year's harvest. Aren't they pretty?

@EthanHolmes @AngelaPreston Wasn’t Glen A Larson LDS? I have a vague recollection of that being the case …

So I'm back on the #JobHunt. If anyone needs a great programmer and software architect with 20+ years of experience, let them know where to find me!

I'm fully remote, have been for many years, have managed international teams, and am fluent in #csharp #rust #typescript #python and more! I speak, read, and write fluent English, a little Spanish, a little German, and a little simplified Mandarin Chinese. I've worked previously in healthcare, insurance, chemical engineering, and marketing.

Talking of organising your home timeline, or filtering, I’ve always treated the home timeline as my main source of info. I always read it, and I always read all of it. Which is why I needed filtering in the first place.

Now this might be fairly obvious to others, but I had a complete blindspot as to the fact that I could not look at the home timeline at all. I could simply create lists for all the people I follow, disregard the home timeline and simply look at the lists only. This would have removed a lot of the pain points I had around filtering …

But that’s not what I did.

And there was a reason for why I did things my way. My personal Fedi client has a notification feature which notifies me when the home timeline has posts. It adds a badge with a count to the app icon. Otherwise, I forget to look at posts at alll and might go through the whole day without reading what is going on in the Fediverse.

Sure, this is another *me* problem 😛 But for my personal reading style, just using lists and ignoring the home timeline would not have worked very well. But I guess if you do not have the specific need that I do, you could simply split up all the people you follow into lists and then disregard the home timeline?

#FediClient #Filtering #Organizing

As is usual with most (any?) simple system, as time goes on, it becomes ever more complicated by accumulating new features 😛

The filtering of the home timeline on my custom Fedi-client started out that way. It was simply a matter of filtering out people who were members of my lists so that my home timeline would be less cluttered.

But then I started following high-volume hashtags. So I decided to add hashtag filtering.

But today I noticed that this also removed low-volume hashtags such as #MachineLearning that I did want to see in my home timeline …

So what do I do to fix that?

This is where the system starts getting complicated 😛 I’m thinking of adding a selective feature which lets me specify which hashtags should be filtered. Sounds simple enough, right? And at it’s core, it is …

But to implement this, I’d have to implement a new UI to show all followed hashtags and add a new property to the hashtag data to indicate that it should be filtered, or not.

I don’t know if I’ll implement this or not. I feel like taking the weekend off from coding. But on the other hand, the urge to code might be so great that I do it anyway. I guess we’ll see …

#FediClient #Filtering #Organizing

New postdoc opportunity in #hcai, people + AI, AI/ethics, and related topics!

We've got a new center at UMD on Values-Centered AI, and want *YOU* to apply for our postdoc position.

Deadline: Mar 31, 2023
Job ad: https://research.umd.edu/sites/default/files/2023-02/vcai-postdoc.pdf
App form: https://docs.google.com/forms/d/e/1FAIpQLSfySV8JY6-aNz6YKW5O-nsi5srvHbDINeBw9LFAhWYkXi1Prw/viewform

This is an opportunity to work with a highly interdisciplinary group in a 1- or 2-year postdoc with no deliverables attached 😉, including connections to HCIL, CML, CLIP, TRACE, SDS Center, Urban Computing, and more!

>

Boosted 7 papers out of 77 in the cs.CV category on arXiv.org today — apparently it’s 7’s all the way 😛

#AI #CV #NewPapers #DeepLearning #MachineLearning

"Data-driven Approach for Automatically Correcting Faulty Road Maps. (arXiv:2211.06544v2 [cs.CV] UPDATED)" — A method to fix faulty road maps (specifically roads displayed on maps) using machine learning.

Paper: http://arxiv.org/abs/2211.06544
Code: https://github.com/soojunghong/image_inpainting_model_for_lane_geomery_discovery

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
GT, Input, and Output denotes t…

"ZITS++: Image Inpainting by Improving the Incremental Transformer on Structural Priors. (arXiv:2210.05950v2 [cs.CV] UPDATED)" — Better image inpainting by detecting structures in the source image using techniques such as edge detection.

Paper: http://arxiv.org/abs/2210.05950
Code: https://github.com/dqiaole/zits_inpainting

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Left (a)-(e): Comparisons of ZI…

"Designing an Encoder for Fast Personalization of Text-to-Image Models. (arXiv:2302.12228v1 [cs.CV])" — A method to teach text-to-image models new concepts in seconds.

Paper: http://arxiv.org/abs/2302.12228

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Our encoder-based method enable…

"Aligning Text-to-Image Models using Human Feedback. (arXiv:2302.12192v1 [cs.LG])" — A fine-tuning method for better aligning generated images to the input text prompt when using diffusion models.

Paper: http://arxiv.org/abs/2302.12192

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
The steps in our fine-tuning me…

"Evaluating the Efficacy of Skincare Product: A Realistic Short-Term Facial Pore Simulation. (arXiv:2302.11950v1 [cs.CV])" — Simulating the effects of skincare products on your skin (specifically the pores) to gauge efficacy of the product.

Paper: http://arxiv.org/abs/2302.11950

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Qualitative results of Facial P…

@at So far I’ve just been ignoring them since it seems to indicate a lack of reading comprehension on their part, or just lack of general interest. But I keep wondering if perhaps I’m being too hard/harsh and maybe they are genuine … My guess is not, but who really knows? 🙂

"Region-Aware Diffusion for Zero-shot Text-driven Image Editing. (arXiv:2302.11797v1 [cs.CV])" — A region-aware text-guided image editing method which aims to replace one entity with another.

What I always wonder with these approaches is whether you can replace a larger entity with a smaller one, or vice versa, (say a horse with a cat) in a way that looks realistic?

Paper: http://arxiv.org/abs/2302.11797
Code: https://github.com/haha-lisa/RDM-Region-Aware-Diffusion-Model

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
The results of the proposed reg…

"Controlled and Conditional Text to Image Generation with Diffusion Prior. (arXiv:2302.11710v1 [cs.CV])" — Using a Diffusion Prior to constrain the generation to a specific domain without altering the larger Diffusion Decoder in a memory and compute efficient way.

Paper: http://arxiv.org/abs/2302.11710

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Diffusion Prior can be trained …

How heat pumps work.

#HomeHeating #HeatPumps

https://www.technologyreview.com/2023/02/14/1068582/everything-you-need-to-know-about-heat-pumps/?utm_source=pocket-newtab

I started reading papers on arXiv.org because I enjoyed learning about generative machine learning models. Then I started boosting papers that I liked because I thought it might be helpful to others …

Now, I sometimes get people asking me to explain the paper to them because they can’t be bothered to read it 😛 What do you do in such a situation?

Would explaining the paper actually do any good? If they can’t be bothered to read the paper, are they going to bother reading my explanation? Is it worthwhile spending the time to explain something to such a person?

I don’t know … It feels as if it’s just a drive-by thing — either because they think it’s funny or because people are like that. Or are there people who are really like “I can’t be bothered to read a paper but if somebody gives me bite-sized info, I’ll take it”?

I think people confuse me 😀

#MachineLearning #Papers #People

A total of 77 papers in the cs.CV category on arXiv.org today — 58 new, 19 updated.

On to the Friday paper reading …

#AI #CV #NewPapers #DeepLearning #MachineLearning

It's small and round, and you rarely find them on the ground.

The wonderful Long-tailed tit.

#nature #naturephotography #birds #birdphotography #birdsofmastodon #photography #miltonkeynes