At this point, “The Last of Us” is the TV show I anticipate the most. Maybe that’s not saying much, but still … 😛

Without spoiling anything, I love how they’ve portrayed zombies and want to learn more about the world with each episode.

We are huge fans of Naughty Dog games given that we played through all of the “Uncharted” games and “The Lost Legacy” too as fast as we could … and loved the story of “Uncharted 4” to bits.

But I didn’t know if we’d like “The Last of Us” since we kind of shy away from killing games and based on the premise, that’s what it sounded like … Now, after starting the TV show, I’m wondering if we should go back and play the game too … Also, there’s apparently a remake and so maybe it really is time to try it?

#TV #Gaming #PlayStation

And that’s a wrap on the start of a new week of paper reading after moving to a new server 🙂

I posted 8 papers out of a total of 51 new and updated papers in the cs.CV category on arXiv.org today.

I must say that the ability to post more than 500 characters on the new server is a huge help since now I don’t have to spend a lot of time cutting down my summary to fit within 500 characters 😛

#AI #CV #NewPapers #DeepLearning #MachineLearning

"Model Complexity-Accuracy Trade-off for a Convolutional Neural Network. (arXiv:1705.03338v1 [cs.CV] CROSS LISTED)" — A study of the model complexity versus accuracy trade-off on MNSIT dataset, providing a concrete framework for handling such a problem, given the worst case accuracy that a system can tolerate.

Paper: http://arxiv.org/abs/1705.03338

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
A graph plotting Accuracy vs Mo…

"MemeTector: Enforcing deep focus for meme detection. (arXiv:2205.13268v2 [cs.CV] UPDATED)" — A methodology that utilizes the visual part of image memes as instances of the regular image class and the initial image memes as instances of the image meme class to force the model to concentrate on the critical parts that characterize an image meme.

Paper: http://arxiv.org/abs/2205.13268
Code: https://github.com/mever-team/memetector

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Example image meme vs. a regula…

@newmanth Yep 🙂 Well, briefly … I couldn’t get it to run locally and there didn’t seem to be that much in the way of docs. Went to the Discord server hoping that the issues I was running into were standard ones and others would have hit them too but nothing on the Discord about those specific issues. So gave up at that point since I was already working with a few other apps 🙂

But will probably go back at some point to see if it would work better the next time around …

@elk

"Learning Sequential Latent Variable Models from Multimodal Time Series Data. (arXiv:2204.10419v2 [cs.LG] UPDATED)" — A self-supervised generative modelling framework to jointly learn a probabilistic latent state representation of multimodal data and the respective dynamics to improve prediction and representation quality.

Paper: http://arxiv.org/abs/2204.10419
Code: https://github.com/utiasstars/visual-haptic-dynamics

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
We learn a sequential latent va…

"REx: Data-Free Residual Quantization Error Expansion. (arXiv:2203.14645v2 [cs.CV] UPDATED)" — A quantization method that leverages residual error expansion, along with group sparsity and an ensemble approximation for better parallelization.

Paper: http://arxiv.org/abs/2203.14645

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Illustration of the proposed me…

"Novel-View Acoustic Synthesis. (arXiv:2301.08730v1 [cs.CV])" — Given the sight and sound observed at a source viewpoint, synthesizing the *sound* of that scene from an unseen target viewpoint using a neural rendering approach.

Paper: http://arxiv.org/abs/2301.08730

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Novel-view acoustic synthesis t…

"Visual Writing Prompts: Character-Grounded Story Generation with Curated Image Sequences. (arXiv:2301.08571v1 [cs.CL])" — A new image-grounded dataset for improving visual story generation due to the fact that existing image sequence collections do not have coherent plots behind them.

Paper: http://arxiv.org/abs/2301.08571

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Comparison of Visual Writing Pr…

"When Source-Free Domain Adaptation Meets Label Propagation. (arXiv:2301.08413v1 [cs.CV])" — An approach that tries to achieve efficient feature clustering from the perspective of label propagation by dividing the target data into inner and outlier samples based on the adaptive threshold of the learning state, and applying a customized learning strategy to best fits the data property.

Paper: http://arxiv.org/abs/2301.08413

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
A toy illustration of target fe…

@mikaelacaron Well, if you have an Apple Silicon device, then a lot of the iOS stuff will run on it … Not Ivory (at least the TestFlight version) and a few others but a lot will but some stuff is annoying — like closing the app will close it fully instead of just keeping it in the dock, which is what I expect from macOS apps 😛

"Open-Set Likelihood Maximization for Few-Shot Learning. (arXiv:2301.08390v1 [cs.CV])" — A generalization of the maximum likelihood principle, in which latent scores down-weighing the influence of potential outliers are introduced alongside the usual parametric model. This implementation can be applied on top of any pre-trained model seamlessly.

Paper: http://arxiv.org/abs/2301.08390

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Intuition behind OSLO. Standard…

@mikaelacaron I thinkI might have like 5 installed but hardly ever use ‘em since I’m no macOS most of the time and there’s sadly very few clients out there that work on #macOS, have the features I really want, *and* work with Akkoma 😛

Was using Ivory to get notifications on iOS (so that I hear when new posts come in) since I have volume turned down on the MacBook, but then I switched to Akkoma and Ivory won’t work with Akkoma 😞

A total of 51 papers in the cs.CV category on arXiv.org today — 28 new, 23 updated.

Time to start reading papers and taking screenshots 😛

#AI #CV #DeepLearning #MachineLearning #NewPapers

Since switching over to the Fedi-side, I have looked at a lot of Mastodon/Fediverse clients …

Outside of the ones I’ve just looked at or used just for a little while, there are a few that I’ve dallied with quite extensively.

First, there’s the Mastonaut fork that I’ve been working on — written in #Swift for #macOS only:
https://github.com/FahimF/Mastonaut

Then, there’s the private fork of IcueCubesApp that I was working on because that app does not work with Akkoma at the moment — this one was #Swift and #SwiftUI but worked on both #iOS and #macOS:
https://github.com/Dimillian/IceCubesApp

Then there was the pure #JavaScript one I started from scratch because I wanted something which would work on any platform.

I stopped working on the JavaScript client because I wanted something that worked now and was too impatient to get my own client working 😛

The IceCubes clone had too many #SwiftUI issues for my taste and so I went back to Mastonaut for the time being. Trouble is, my wife uses iOS for most of her Fediverse traversal and so that leaves her out of things …

I was hoping to get Mastonaut working to my taste so that I have something I can use now and then go back to the JavaScript client so that I have an option which works anywhere. And I thought I was on my way there …

But today I discover that Mastonaut will not reblog when connected to Akkoma. And based on the bug report below, it looks as if this is an Akkoma/Pleroma issue since the error I get is the exact same one:
https://github.com/TwidereProject/Twidere-Android/issues/1460

I’d really like to have one client which works on every platform, has all the features that make life easy for me, and works with all the different Fediverse servers. Is that too much to ask for? 😛

#Fediverse #Software #Apps

It's #Sunday and I can't hear a thing so it must be #SilentSunday

I'm sneaking this in for #StandingStoneSunday just because I like it.

I always think of cairns and Trig Points as some sort of #StoneCircle where you have to go out and join the dots.

#Photography #Cumbria #LakeDistrict

Follow up #introduction since I didn’t include actual searchable tags! In real life I like jokes and roller derby but here I mainly talk about #nlproc #nlp #scienceofdl #trainingdynamics #deeplearning #losslandscapes #compositionality #generalization

Looking to connect socially and professionally with women and nonbinary folks in iOS! Especially if you’re brown, queer, genderqueer, or otherwise marginalized.

Please boost for reach 💜

#SilentSunday