Streaming the video from raspberry pi to the internet at https://t.co/MWrcd1d9cK, california’s outlook It’s down now, was very interesting though

We introduce a training-free, critique-and-correct multi-agent framework for semantic segmentation that enforces adherence to long and detailed textual labeling guidelines.

Excited to share that our work “Guideline-Consistent Segmentation via Multi-Agent Refinement” has been accepted to AAAI 2026 (Main Technical Track)! 🎊 Arxiv preprint: https://t.co/uaTdZrZhM5 #AAAI2026 #ArtificialIntelligence #ComputerVision #MultiAgent

I wish I could develop taste like Trader Joe, inspirational work,

Quite impressed with agents for first time, generated my new website in next.js with most of the things written by agent in vs code: https://t.co/MhDvGhOl7E Previous version that I tried to align content wise:

Blog: https://t.co/eoo9Rhjuj7 Post on Linear Probing for evaluating the quality of learned representations via self-supervised learning on Masked Image Prediction on MNIST.

I try to describe how training objective can be defined in a way to learn relevant capabilities in absence of labelled data.

Post on Masked Image Prediction on MNIST utilizing self-supervised learning. SSL is backbone for large foundational models like DINO, Sonata(3D point cloud) etc Blog: https://t.co/2Cm4Ulm3DA Demo:

Reading the SLAM handbook written by frontier robotics people, wrote a blog post on their views on past, present and future of SLAM and my reflection: https://t.co/NbLsMgSSaN Handbook:

3d gaussian represented as triangle fan with alpha falloff is a neat idea

Post on sorting algorithm(merge sort) on CPU vs GPU(CUDA) with comparisons: https://t.co/Jn6qVvsqIb

Post of gaussian splatting first iteration: https://t.co/jC3onZ9sah and can be visualized here:

Post on sorting algorithms reiterated:

It works on planar surfaces where one can find point correspondences, calculate homography, and finally apply inverse homography to undistort. More results are here:

In Santa Cruz, winter is coming! Given that the water is cold and the tides are high, warning signs are important. But how can you see them without the right view? I get some interesting results from Hartley & Zisserman on undistorting perspective distortions shown below.

Hi lilly!! 🩵

We ordered stickers for many Julia orgs for @JuliaConOrg. See you there! #julialang #juliacon

Interested in learning image processing in Julia? Come Join us Tomorrow at JuliaCon'23 Presenters: Tim Holy and Ashwani Rathee Date Time, Venue: Tuesday 25th July 9AM EST, MIT Stata Center(32-G449 (Kiva)) Notebooks: https://t.co/LpE99JwQbS #julia #julialang #imageprocessing

Images.jl v0.26 is out! Adds major upgrades with changes in ImageMorphology.jl, and now re-exports ImageBinarization.jl and ImageCorners.jl. More details here: https://t.co/xyIiwHhYbe #julia #julialang #computervision #images

ImageQualityIndexes v0.3.6 is out Since the last announcement, several things have changed: - Multi Scale SSIM was added - Precompilation support - Assess SSIM working on n-dimensional arrays - Lazy loading of large dependencies Repository: https://t.co/WwhTHrHWD1.