indoor positioning @IndoorLocalize

Introducing D4RT: A unified AI model for 4D scene reconstruction and tracking across space and time. 🎯 Catch the demo with Skanda Koppula at 12 pm at our #CVPR2026 Google booth kiosk! d4rt-paper.github.io @GoogleDeepMind

18 138 1K 78K 708

View Details

Sven Elflein @s_elflein

2 months ago

We just released code and model! Go check it out! Code: github.com/nv-dvl/vgg-ttt Model: huggingface.co/nvidia/vgg-ttt

Sven Elflein @s_elflein

5 months ago

🚀 Exciting news! We’re introducing VGG-T³: a scalable model for offline feed-forward 3D reconstruction that finally tackles the "quadratic bottleneck." Ever wanted to have VGGT reconstruct a 1,000-image scene in seconds instead of 10 minutes and use it for visual localization?

7 78 548 85K 350

9 63 467 46K 367

View Details

Massimiliano Viola @massiviola01

4 months ago

Then, they process the images with pixel-aligned generative models to synthesize versions in thermal, night, and depth formats. This step transforms the established matches into multiple modalities. The result is a massive dataset of 800 million synthetic image pairs. Dump these into SOTA marching models, retrain, and boom, you've got the most universal image matcher out there! 🚀 MatchAnything is genuinely an insane framework, generalizing to many modalities well beyond its training set. Simply the confirmation that scaling data is always one way to go :) Paper: arxiv.org/abs/2501.07556 Website: zju3dv.github.io/MatchAnything/

0 2 12 929 13

View Details

Qiusheng Wu @giswqs

4 months ago

The GeoAI Python package now supports object detection using pre-trained models from the GeoDeep libarary (github.com/uav4geo/GeoDeep). The supported object types include cars, trees, birds, planes, aerovision, utilities, buildings, and roads. Try it out: GitHub: github.com/opengeos/geoai Notebook example: opengeoai.org/examples/geode… #geospatial #geoai #opensource #python

12 168 1K 87K 1K

View Details

indoor positioning @IndoorLocalize

4 months ago

@TrbimixsQ @howie_serious 候补好久了，还没开放

0 0 1 146 0

View Details

AIGCLINK @aigclink

5 months ago

Karpathy深夜炸出自动化AI研究新项目：autoresearch ，让AI帮你做自动化循环LLM训练，你给指令然后躺着等结果 5分钟即完成一次训练，单卡能跑人工不用在苦哈哈的去改Python代码、调参数、改神经网络架构，只给Prompt，AI自己去修改代码优化模型给AIAgent一个真实的LLM训练环境，让它通宵达旦自主进行实验研究，人类给Prompt来指导AI思考和研究 Karpathy把这个项目打包成一个独立的迷你代码库，方便上手体验，这个项目本质上是nanochat模型训练核心的精简版，被压缩成一个630行的单文件代码，能在单GPU上运行然后人训练LLM就变成了AI训练LLM 你告诉AI：去改改训练代码，看能不能让模型效果更好 AI收到指令：修改train.py，跑5分钟训练，看结果如果结果变好，保留修改如果结果变差，丢弃重新改然后重复循环…… 这个项目里可以看到AI自主做研究的可行性，AI不仅写代码已经能开始接管最硬核的AI算法研究本身了或许谁能写出更好的Prompt，谁能调教出更聪明的AI智能体才是核心竞争力，在这个范式中，Prompt就是新的源代码 #autoresearch #AI训练LLM #AI自主研究

Andrej Karpathy @karpathy

5 months ago

I packaged up the "autoresearch" project into a new self-contained minimal repo if people would like to play over the weekend. It's basically nanochat LLM training core stripped down to a single-GPU, one file version of ~630 lines of code, then: - the human iterates on the

1K 4K 28K 11.1M 39K

8 47 184 31K 208

View Details

Suhas Kotha @kothasuhas

5 months ago

to improve fine-tuning data efficiency, replay generic pre-training data not only does this reduce forgetting, it actually improves performance on the fine-tuning domain! especially when fine-tuning data is scarce in pre-training (w/ @percyliang)

15 64 498 74K 377

View Details

ǝɥǝ⅄ @YeheLiu

5 months ago

Forgot to share this: latest COLMAP (v3.14) and pycolmap built for all systems and configs. Aliked + LightGlue is now natively supported. Note: cuDSS only adds 20% mapping speed and is unstable in some systems. github.com/lyehe/build_gp…

2 7 59 4K 50

View Details

Marshall Richards @marshallrichrds

5 months ago

@cryptoleek I have some instructions on my repo: github.com/marshallrichar… Will update more tomorrow.

4 5 56 2K 78

View Details

Luigi Freda @LuigiFreda

5 months ago

A new major release of #pySLAM is out ✨ This update pushes the framework forward on: • Runtime performance →new C++ core • #SemanticSLAM tightly coupled with volumetric mapping • End-to-end multi-view 3D scene inference 🧵 Thread below 🔗github.com/luigifreda/pys…