"Learning on tree architectures outperforms a convolutional feedforward network. (arXiv:2211.11378v3 [cs.CV] UPDATED)" — A 3-layer tree architecture inspired by experimental-based dendritic tree adaptations is developed and applied to the offline and online learning of the CIFAR-10 database to show that this architecture outperforms the achievable success rates of the 5-layer convolutional LeNet.

Paper: http://arxiv.org/abs/2211.11378

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Comparison of offline and onlin…

"Continual Learning by Modeling Intra-Class Variation. (arXiv:2210.05398v2 [cs.LG] UPDATED)" — An examination of memory-based continual learning which identifies that large variation in the representation space is crucial for avoiding catastrophic forgetting.

Paper: http://arxiv.org/abs/2210.05398
Code: https://github.com/yulonghui/moca

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Average intra-class angle devia…

"Scalable and Equivariant Spherical CNNs by Discrete-Continuous (DISCO) Convolutions. (arXiv:2209.13603v3 [cs.CV] UPDATED)" — A hybrid discrete-continuous (DISCO) group convolution for spherical convolutional neural networks (CNN) that is simultaneously equivariant and computationally scalable to high-resolution.

Paper: http://arxiv.org/abs/2209.13603

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Spherical CNN categorization

"Multi-Level Visual Similarity Based Personalized Tourist Attraction Recommendation Using Geo-Tagged Photos. (arXiv:2109.08275v2 [cs.MM] UPDATED)" — A geo-tagged photo based tourist attraction recommendation system which utilizes the visual contents of photos and interaction behavior data to obtain the final embeddings of users and tourist attractions, which are then used to predict the visit probabilities.

Paper: http://arxiv.org/abs/2109.08275
Code: https://github.com/revaludo/MEAL

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
An illustration of the multi-le…

"BiAdam: Fast Adaptive Bilevel Optimization Methods. (arXiv:2106.11396v3 [math.OC] UPDATED)" — A novel fast adaptive bilevel framework to solve stochastic bilevel optimization problems that the outer problem is possibly nonconvex and the inner problem is strongly convex.

Paper: http://arxiv.org/abs/2106.11396

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
The basic idea of the convergen…

"Sparse Oblique Decision Trees: A Tool to Understand and Manipulate Neural Net Features. (arXiv:2104.02922v2 [cs.LG] UPDATED)" — An effort to understanding which of the internal features computed by the neural net are responsible for a particular class, by mimicking part of the neural net with an oblique decision tree having sparse weight vectors at the decision nodes.

Paper: http://arxiv.org/abs/2104.02922

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Mimicking part of a neural net …

"Don't Play Favorites: Minority Guidance for Diffusion Models. (arXiv:2301.12334v1 [cs.LG])" — A framework that can make the generation process of the diffusion models focus on the minority samples, which are instances that lie on low-density regions of a data manifold.

Paper: http://arxiv.org/abs/2301.12334

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Diffusion models play favorites…

"SEGA: Instructing Diffusion using Semantic Dimensions. (arXiv:2301.12247v1 [cs.CV])" — A semantic guidance method for diffusion models to allow making subtle and extensive edits and changes in composition and style, as well as optimize the overall artistic conception.

Paper: http://arxiv.org/abs/2301.12247

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Semantic control over image gen…

@at LOL. I have to admit that I posted about that one partially because of the title. I assume that they selected that title on purpose because the alternative is that they have no idea what that means and that would be a … self-fulfilling title? 😛

"Anticipate, Ensemble and Prune: Improving Convolutional Neural Networks via Aggregated Early Exits. (arXiv:2301.12168v1 [cs.LG])" — A new training technique based on weighted ensembles of early exits, which aims at exploiting the information in the structure of networks to maximise their performance.

Paper: http://arxiv.org/abs/2301.12168

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Outline of the AEP technique. O…

@AngelaPreston Thank you 🙂 My trouble usually is that once I start thinking about writing, the characters start taking over and I get so many scenes playing out in my head that there’s no way to turn it off … I wish they had a brain - computer interface so that I can just sit there and let the computer transcribe everything but alas, I guess I’ll just have to type it all out myself 😛

"ClusterFuG: Clustering Fully connected Graphs by Multicut. (arXiv:2301.12159v1 [cs.CV])" — A simpler and potentially better performing graph clustering formulation based on multicut (a.k.a. weighted correlation clustering) on the complete graph.

Paper: http://arxiv.org/abs/2301.12159

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Example illustration of dense m…

"Towards Equitable Representation in Text-to-Image Synthesis Models with the Cross-Cultural Understanding Benchmark (CCUB) Dataset. (arXiv:2301.12073v1 [cs.CV])" — A culturally-aware priming approach for text-to-image synthesis using a small but culturally curated dataset to fight the bias prevalent in giant datasets.

Paper: http://arxiv.org/abs/2301.12073

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Sample images generated for fiv…

"Cross-Architectural Positive Pairs improve the effectiveness of Self-Supervised Learning. (arXiv:2301.12025v1 [cs.CV])" — A novel self-supervised learning approach that leverages Transformer and CNN simultaneously to overcome the issues with existing self-supervised techniques which have extreme computational requirements and suffer a substantial drop in performance with a reduction in batch size or pretraining epochs.

Paper: http://arxiv.org/abs/2301.12025

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
In our proposed self-supervised…

"Improved knowledge distillation by utilizing backward pass knowledge in neural networks. (arXiv:2301.12006v1 [cs.LG])" — Addressing the issue with Knowledge Distillation (KD) where there is no guarantee that the model would match in areas for which you do not have enough training samples, by generating new auxiliary training samples based on extracting knowledge from the backward pass of the teacher in the areas where the student diverges greatly from the teacher.

Paper: http://arxiv.org/abs/2301.12006

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
a) Minimization Step: Using the…

"RGB Arabic Alphabets Sign Language Dataset. (arXiv:2301.11932v1 [cs.CV])" — An Arabic Alphabet Sign Language (AASL) dataset comprising of 7,856 raw and fully labelled RGB images of the Arabic sign language alphabets which might be the first such publicly available dataset.

Paper: http://arxiv.org/abs/2301.11932

#AI #CV #NewPaper #DeepLearning #MachineLearning

<<Find this useful? Please boost so that others can benefit too 🙂>>
Sample images from the dataset

A total of 89 papers in the cs.CV category on arXiv.org today — 37 new, 52 updated.

Lots of updates today ... So on to seeing if there's anything interesting in there 🙂

#AI #CV #NewPapers #DeepLearning #MachineLearning

The worst thing about having your password stolen is having to rename the dog.

@Miteroni Thanks for the encouragement 🙂 I’m fairly (80%-ish?) sure that I’ll do it, just need to get the logistics together and then actually start writing … Hopefully, this weekend.

@davemark Personally, I want everything I build to work on both macOS and iOS since if I use it, then my wife will want to use it and she’ll generally be using it from her iPad. So SwiftUI is the preferred option for me most of the time these days.

Went through several Mastodon clients because of this actually 🙂 I liked Mastonaut because the source was available and it was native macOS, but then my wife wanted the features I was building. Looked at at least one which was pure SwiftUI and modified it a bunch but kept hitting issues getting certain things to work — like the UITextView thing I mentioned previously.

Now I’m using a Swift app (with some SwiftUI code) which was mainly written as an iOS app with Catalyst support and I’m modifying the Catalyst part to work for my needs on a Mac. Working out OK so far but it’s a bit of a struggle with Catalyst.

So based on this particular experience, developing for iOS with Catalyst support for macOS would seem to be the the way to go … But my heart is set on trying to get pure native apps on both sides of the fence if I can and so I’d really want it to be SwiftUI cross platform. But that takes a lot more work … and I’m lazy 😛