# Point cloud

A set of data points in 3D space, typically containing XYZ coordinates and possibly attributes such as intensity, color, or classification.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Efficient Genotype Classification from 3D MRI in ADPKD via Density-Aware Point Sampling and E(3)-invariance

DevFeed: [Efficient Genotype Classification from 3D MRI in ADPKD via Density-Aware Point Sampling and E(3)-invariance](<https://devfeed.tech/articles/efficient-genotype-classification-from-3d-mri-in-adpkd-via-density-aware-point-sampling-and-e-3-invariance-135824.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.17758>)

Author: Kwanghee Lee, Hyeongsub Lee, Youngwoo Kim, Hyunji Jeong

Published: 2026-11-16T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [Point cloud](<https://devfeed.tech/topics/point-cloud.md>), [image recognition](<https://devfeed.tech/topics/image-recognition.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [canonicalization](<https://devfeed.tech/tags/canonicalization.md>), [classification](<https://devfeed.tech/tags/classification.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [data-efficiency](<https://devfeed.tech/tags/data-efficiency.md>), [dataset](<https://devfeed.tech/tags/dataset.md>), [deep-learning](<https://devfeed.tech/tags/deep-learning.md>), [deep-learning-models](<https://devfeed.tech/tags/deep-learning-models.md>), [detection](<https://devfeed.tech/tags/detection.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [faster](<https://devfeed.tech/tags/faster.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [gpu-memory](<https://devfeed.tech/tags/gpu-memory.md>), [invariance](<https://devfeed.tech/tags/invariance.md>), [point](<https://devfeed.tech/tags/point.md>), [sampling](<https://devfeed.tech/tags/sampling.md>)

### AI overview

E3PN classifies five ADPKD genotypes from 3D MRI using a point representation to reduce memory use, density-aware sampling across both kidneys, and coordinate canonicalization to incorporate E(3)-invariance. On the authors' dataset, it trained 163 times faster and used 103 times less GPU memory than a 3D ResNet, while improving F1 from 0.165 to 0.226 over both baselines. The abstract presents this as a first benchmark; it does not provide evidence about performance beyond that dataset.

### Source excerpt

arXiv:2611.17758v1 Announce Type: cross Abstract: Early detection of genetic kidney diseases, especially autosomal dominant polycystic kidney disease (ADPKD), is crucial because certain genotypes such as truncating PKD1 mutations cause rapid kidney growth and early progression to end-stage kidney disease. MRI-based genotype inference is a potential low-cost complement to genetic testing. However, although many deep learning methods have been developed for classifying various MR images, applying them to ADPKD remains challenging: (1) abdominal MR images are generally too large to be handled by deep learning models designed for smaller images like brain MRI, and (2) the amount of labeled data for ADPKD genetic mutations is insufficient to train deep learning models. In order to address these challenges, we propose E3PN, an efficient 3D MRI-based ADPKD genotype classification network that relies on point sampling and E(3)-invariance. We utilize a point representation of 3D MRI, which consumes less GPU memory than a voxel representation. For further memory efficiency, 3D points are sampled from a mixture distribution over the two kidneys in each image. Since E(3)-invariance has been shown to improve data efficiency, we also incorporate it into our model through coordinate canonicalization. On our ADPKD dataset, E3PN trains 163 times faster and uses 103x less GPU memory than a 3D ResNet, while also improving F1 from 0.165 to 0.226 over both baselines. These results establish a first benchmark for five-way ADPKD genotype classification from 3D MRI and show that point sampling and coordinate canonicalization yield two-orders-of-magnitude efficiency gains with a modest but consistent improvement in F1.

## A Google engineer points 48 AI agents at NASA Lidar and finds an unmapped Maya pyramid

DevFeed: [A Google engineer points 48 AI agents at NASA Lidar and finds an unmapped Maya pyramid](<https://devfeed.tech/articles/a-google-engineer-points-48-ai-agents-at-nasa-lidar-and-finds-an-unmapped-maya-pyramid-129512.md>)

Original publisher: [Read original article](<https://blog.adafruit.com/2026/11/11/a-google-engineer-points-48-ai-agents-at-nasa-lidar-and-finds-an-unmapped-maya-pyramid/>)

Author: Anne Barela

Published: 2026-11-11T21:27:39Z

Content type: news

Language: en

Sources: [Adafruit Industries](<https://devfeed.tech/sources/adafruit-industries-makers-hackers-artists-designers-and-engineers.md>)

Topics: [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>), [Point cloud](<https://devfeed.tech/topics/point-cloud.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [data](<https://devfeed.tech/tags/data.md>), [engineer](<https://devfeed.tech/tags/engineer.md>), [google](<https://devfeed.tech/tags/google.md>), [history](<https://devfeed.tech/tags/history.md>), [lidar](<https://devfeed.tech/tags/lidar.md>), [maps](<https://devfeed.tech/tags/maps.md>), [maya](<https://devfeed.tech/tags/maya.md>), [nasa](<https://devfeed.tech/tags/nasa.md>), [public](<https://devfeed.tech/tags/public.md>), [science](<https://devfeed.tech/tags/science.md>), [shapes](<https://devfeed.tech/tags/shapes.md>), [site](<https://devfeed.tech/tags/site.md>), [software](<https://devfeed.tech/tags/software.md>), [software-engineer](<https://devfeed.tech/tags/software-engineer.md>)

### AI overview

Using public NASA lidar data, a software engineer had 48 AI agents inspect images from about 1,200 km² of lidar strips, following a written guide rather than a special model or classifier. They flagged a hilltop site in Mexico with features including an oval pyramid estimated at 8-12 m tall, then checked the raw point cloud to confirm the shapes were on the ground. The engineer drafted a Spanish report for INAH, which replied that it had no registered site at that location. The work took under 72 hours. A later update says the site had already been mapped in a 2025 paper the agents missed, making this a rediscovery rather than a new find.

### Source excerpt

A software engineer, writing at pietromiotti.com, used AI agents to go through public NASA lidar data and turn up a Maya site in Mexico that wasn't in the Mexican government's register. Lidar fires laser pulses through gaps in the jungle canopy and maps the ground underneath, so structures hidden by trees show up as shapes [...]

## Do We Still Need Rotation-Invariant Architectures? A Statistical Robustness Benchmark for Point Cloud Classification

DevFeed: [Do We Still Need Rotation-Invariant Architectures? A Statistical Robustness Benchmark for Point Cloud Classification](<https://devfeed.tech/articles/do-we-still-need-rotation-invariant-architectures-a-statistical-robustness-benchmark-for-point-cloud-classification-116868.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.03952>)

Author: Abdulla Nasir Chowdhury, Gulpi Qorik Oktagalu Pratamasunu, Kazuhiro Fukui

Published: 2026-11-04T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Benchmark](<https://devfeed.tech/topics/benchmark.md>), [Point cloud](<https://devfeed.tech/topics/point-cloud.md>), [enhancing AI model accuracy](<https://devfeed.tech/topics/enhancing-ai-model-accuracy.md>)

Tags: [benchmark](<https://devfeed.tech/tags/benchmark.md>), [classification](<https://devfeed.tech/tags/classification.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [invariant](<https://devfeed.tech/tags/invariant.md>), [point](<https://devfeed.tech/tags/point.md>), [robustness](<https://devfeed.tech/tags/robustness.md>), [rotation](<https://devfeed.tech/tags/rotation.md>)

### AI overview

Across five domains, rotation-invariant (RI) point-cloud architectures had the strongest pose robustness but were among the weakest on noise robustness in four domains. Adding an embedding invariance loss to the ten architectures improved the corruption types targeted by its perturbations when baseline performance had room to improve, including mean accuracy gains of 10.5 percentage points for occlusion and 13.4 for lidar. On ModelNet40-C, the best standard backbone and best RI architecture performed within 2.5 percentage points of each other on unseen corruptions across every seed at matched training budgets. Invariance-trained RIConv++ was the most seed-stable top model. With wider rotation augmentation, standard backbones had a flat accuracy offset under full-sphere rotation. The authors report releasing the benchmark and a package for robustness analysis and reproduction.

### Source excerpt

arXiv:2611.03952v1 Announce Type: cross Abstract: A substantial line of work builds rotation invariance directly into point cloud architectures. We benchmark 10 architectures (PointNet, PointNet++, DGCNN, RIConv, RIConv++, RISurConv, PointMLP, PointTransformer, and the equivariant VN-PointNet and VN-DGCNN) across 5 domains under pose, noise, and occlusion, and analyze the results with confirmatory hypothesis tests rather than point estimates. The three rotation-invariant (RI) architectures achieve the best pose robustness but rank among the lowest on noise robustness in 4 of 5 domains. We ask whether this cost is intrinsic to invariance or specific to how invariance is obtained. We retrain all 10 architectures unchanged, with a simple invariance loss on the embedding and three training seeds each, and evaluate all models on the independently designed ModelNet40-C. Invariance training transfers narrowly but reliably. It improves exactly the corruption families its perturbations target wherever the baseline has headroom (mean occlusion +10.5, lidar +13.4 percentage points of accuracy). At matched training budget the best standard backbone and the best RI architecture end on par on unseen corruptions (within 2.5 percentage points in every seed), and invariance-trained RIConv++ is the most seed-stable top model. Once rotation augmentation is widened, standard backbones concede only a flat, angle-independent accuracy offset under full-sphere rotation. We release the benchmark along with a package for analyzing point cloud robustness and reproducing our results.

## Supporter spotlight: Jochen Sprickerhof on reproducible builds

DevFeed: [Supporter spotlight: Jochen Sprickerhof on reproducible builds](<https://devfeed.tech/articles/supporter-spotlight-jochen-sprickerhof-on-reproducible-builds-34163.md>)

Original publisher: [Read original article](<https://reproducible-builds.org/news/2026/09/15/supporter-spotlight-jochen-sprickerhof/>)

Published: 2026-09-15T03:54:00Z

Content type: news

Language: en

Sources: [Reproducible Builds](<https://devfeed.tech/sources/reproducible-builds-org.md>)

Topics: [reproducible builds](<https://devfeed.tech/topics/reproducible-builds.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>), [Debian](<https://devfeed.tech/topics/debian.md>), [Programming](<https://devfeed.tech/topics/programming.md>), [F-Droid](<https://devfeed.tech/topics/f-droid.md>), [Robotics](<https://devfeed.tech/topics/robotics.md>), [Point cloud](<https://devfeed.tech/topics/point-cloud.md>)

Tags: [debian](<https://devfeed.tech/tags/debian.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [org](<https://devfeed.tech/tags/org.md>), [programming](<https://devfeed.tech/tags/programming.md>), [reproducible-builds](<https://devfeed.tech/tags/reproducible-builds.md>), [robotics](<https://devfeed.tech/tags/robotics.md>)

### AI overview

An interview with Jochen Sprickerhof, a freelance programmer and Reproducible Builds core team member, covering his work on Debian, F-Droid, software projects, and bit-for-bit reproduction of Debian packages.

### Source excerpt

The Reproducible Builds project relies on several projects, supporters and sponsors for financial support, but they are also valued as ambassadors who spread the word about our project and the work that we do. This is the ninth installment in a series featuring the projects, companies and individuals who support the Reproducible Builds project. We started this series by featuring the Civil Infrastructure Platform project, and followed this up with a post about the Ford Foundation as well as recent ones about ARDC, the Google Open Source Security Team (GOSST), Bootstrappable Builds, the F-Droid project, David A. Wheeler, Simon Butler and Kees Cook. Today, however, we will be talking with Jochen Sprickerhof, one of the newer members of the Reproducible Builds project core team. Vagrant Cascadian: Could you tell me a bit about yourself? What sort of things do you work on? Jochen Sprickerhof: I am a freelance programmer working on Open Source. Mainly doing Debian, F-Droid and some smaller software projects. In general I made it a habit to look into every software I use and try to fix bugs or add features I need. In Debian, I maintain about 180 packages with topics covering home banking, build systems and robotics. Most of my time, I currently work on reproduce.debian.net, where we try to bit-for-bit reproduce the packages distributed by Debian. Vagrant: Could you describe the path that lead you to working on reproducible builds? Jochen: I started my Debian journey as a teenager, converting my school to Debian and serving as its system administrator for 13 years. After studying Applied System Science, I joined the university's robotics labs, where I worked on the Robot Operating System (ROS) and the Point Cloud Library (PCL). In the end, I enjoyed programming more than writing papers, so I eventually left academia for a robotics startup. Some years ago, I realized that the open source work I was doing in my spare time was actually the work I cared most about. Nowadays I

## AI-generated synthetic neurons speed up brain mapping

DevFeed: [AI-generated synthetic neurons speed up brain mapping](<https://devfeed.tech/articles/ai-generated-synthetic-neurons-speed-up-brain-mapping-6748.md>)

Original publisher: [Read original article](<https://research.google/blog/ai-generated-synthetic-neurons-speed-up-brain-mapping/>)

Published: 2026-04-16T12:18:00Z

Content type: article

Language: en

Sources: [The latest research from Google](<https://devfeed.tech/sources/the-latest-research-from-google.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [Google](<https://devfeed.tech/topics/google.md>), [Point cloud](<https://devfeed.tech/topics/point-cloud.md>), [Training AI Models](<https://devfeed.tech/topics/training-ai-models.md>), [dataset](<https://devfeed.tech/topics/dataset.md>), [neuron](<https://devfeed.tech/topics/neuron.md>), [data](<https://devfeed.tech/topics/data.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [accelerate](<https://devfeed.tech/tags/accelerate.md>), [ai](<https://devfeed.tech/tags/ai.md>), [classification](<https://devfeed.tech/tags/classification.md>), [errors](<https://devfeed.tech/tags/errors.md>), [general-science](<https://devfeed.tech/tags/general-science.md>), [generation](<https://devfeed.tech/tags/generation.md>), [google](<https://devfeed.tech/tags/google.md>), [health-bioscience](<https://devfeed.tech/tags/health-bioscience.md>), [iclr](<https://devfeed.tech/tags/iclr.md>), [iclr-2026](<https://devfeed.tech/tags/iclr-2026.md>), [images](<https://devfeed.tech/tags/images.md>), [machine-intelligence](<https://devfeed.tech/tags/machine-intelligence.md>), [mapping](<https://devfeed.tech/tags/mapping.md>), [model](<https://devfeed.tech/tags/model.md>), [models](<https://devfeed.tech/tags/models.md>), [neuron](<https://devfeed.tech/tags/neuron.md>), [partners](<https://devfeed.tech/tags/partners.md>), [research](<https://devfeed.tech/tags/research.md>), [scale](<https://devfeed.tech/tags/scale.md>), [science](<https://devfeed.tech/tags/science.md>), [synthetic](<https://devfeed.tech/tags/synthetic.md>), [visualization](<https://devfeed.tech/tags/visualization.md>)

### AI overview

Google Research describes how MoGen generates synthetic neuronal shapes to improve AI models that reconstruct brain wiring maps. Adding synthetic training examples reduced reconstruction errors by 4.4%, potentially saving 157 person-years of manual proofreading for a complete mouse brain.

### Source excerpt

General Science

## Metric Depth from a Single Camera: How Robots See Without LiDAR

DevFeed: [Metric Depth from a Single Camera: How Robots See Without LiDAR](<https://devfeed.tech/articles/metric-depth-from-a-single-camera-how-robots-see-without-lidar-39657.md>)

Original publisher: [Read original article](<https://www.gauravsarma.com/posts/2026-04-07_metric-depth-from-a-single-camera>)

Published: 2026-04-07T00:00:00Z

Content type: tutorial

Language: en

Sources: [Gaurav Sarma's](<https://devfeed.tech/sources/gaurav-sarma-s-blog.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Algorithms](<https://devfeed.tech/topics/algorithms.md>), [systems](<https://devfeed.tech/topics/systems.md>), [Point cloud](<https://devfeed.tech/topics/point-cloud.md>), [3D](<https://devfeed.tech/topics/3d.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [ai](<https://devfeed.tech/tags/ai.md>), [algorithm](<https://devfeed.tech/tags/algorithm.md>), [camera](<https://devfeed.tech/tags/camera.md>), [lidar](<https://devfeed.tech/tags/lidar.md>)

### AI overview

This article explains how robots can estimate dense, metric depth from a single camera by combining an AI depth-estimation model with a tracking algorithm that provides real-world scale. It contrasts this approach with LiDAR and describes the role of depth ambiguity in camera-based navigation.

### Source excerpt

Take a photo of a coffee mug on your desk. Now look at that photo...

## End-to-end scalable vision intelligence pipeline using LIDAR 3D Point Clouds on AWS

DevFeed: [End-to-end scalable vision intelligence pipeline using LIDAR 3D Point Clouds on AWS](<https://devfeed.tech/articles/end-to-end-scalable-vision-intelligence-pipeline-using-lidar-3d-point-clouds-on-aws-50145.md>)

Original publisher: [Read original article](<https://aws.amazon.com/blogs/hpc/end-to-end-scalable-vision-intelligence-pipeline-using-lidar-3d-point-clouds-on-aws/>)

Author: Frantz Lohier

Published: 2026-01-27T18:48:29Z

Content type: tutorial

Language: en

Sources: [AWS HPC Blog](<https://devfeed.tech/sources/aws-hpc-blog.md>)

Topics: [Amazon Web Services](<https://devfeed.tech/topics/aws.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Point cloud](<https://devfeed.tech/topics/point-cloud.md>), [cloud-infrastructure](<https://devfeed.tech/topics/cloud-infrastructure.md>), [Photogrammetry](<https://devfeed.tech/topics/photogrammetry.md>), [Algorithms](<https://devfeed.tech/topics/algorithms.md>), [Serverless](<https://devfeed.tech/topics/serverless.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [aws](<https://devfeed.tech/tags/aws.md>), [best-practices](<https://devfeed.tech/tags/best-practices.md>), [high-performance-computing](<https://devfeed.tech/tags/high-performance-computing.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [lidar](<https://devfeed.tech/tags/lidar.md>), [pipeline](<https://devfeed.tech/tags/pipeline.md>), [point](<https://devfeed.tech/tags/point.md>), [scale](<https://devfeed.tech/tags/scale.md>), [slam](<https://devfeed.tech/tags/slam.md>), [technical-how-to](<https://devfeed.tech/tags/technical-how-to.md>), [vision](<https://devfeed.tech/tags/vision.md>)

### AI overview

This tutorial describes an AWS-based pipeline for analyzing LiDAR 3D point clouds and interpreting complex scenes with AI. It covers SLAM, photogrammetry, point-cloud reconstruction and image or video interpretation, with a focus on scaling compute-intensive algorithms and processing large historical datasets using serverless AWS services.

### Source excerpt

Elevate your mining and construction business to new heights with the power of 3D world modeling and AI-driven scene interpretation. Our latest blog post delves into the underlying technologies, from SLAM and photogrammetry to point-cloud analysis, and demonstrates how to scale the execution of these compute-intensive algorithms on AWS. Unlock the potential to optimize your workflows, improve decision-making, and drive sustainable growth.

## The Čech Complex and the Vietoris-Rips Complex

DevFeed: [The Čech Complex and the Vietoris-Rips Complex](<https://devfeed.tech/articles/the-cech-complex-and-the-vietoris-rips-complex-40386.md>)

Original publisher: [Read original article](<https://www.jeremykun.com/2015/08/06/cech-vietoris-rips-complex/>)

Published: 2015-08-06T09:00:00Z

Content type: article

Language: en

Sources: [Jeremy Kun](<https://devfeed.tech/sources/jeremy-kun.md>)

Topics: [data](<https://devfeed.tech/topics/data.md>), [dataset](<https://devfeed.tech/topics/dataset.md>), [Point cloud](<https://devfeed.tech/topics/point-cloud.md>), [Computing](<https://devfeed.tech/topics/computing.md>)

Tags: [approximation](<https://devfeed.tech/tags/approximation.md>), [cech-complex](<https://devfeed.tech/tags/cech-complex.md>), [complex](<https://devfeed.tech/tags/complex.md>), [computational-topology](<https://devfeed.tech/tags/computational-topology.md>), [data-mining](<https://devfeed.tech/tags/data-mining.md>), [data-science](<https://devfeed.tech/tags/data-science.md>), [dataset](<https://devfeed.tech/tags/dataset.md>), [homology](<https://devfeed.tech/tags/homology.md>), [linear-algebra](<https://devfeed.tech/tags/linear-algebra.md>), [mathematics](<https://devfeed.tech/tags/mathematics.md>), [persistent-homology](<https://devfeed.tech/tags/persistent-homology.md>), [point](<https://devfeed.tech/tags/point.md>), [points](<https://devfeed.tech/tags/points.md>), [simplicial-complex](<https://devfeed.tech/tags/simplicial-complex.md>), [vietoris-rips-complex](<https://devfeed.tech/tags/vietoris-rips-complex.md>)

### AI overview

This article introduces computational topology for analyzing the shape of data. It explains how point clouds can be converted into simplicial complexes so homology and persistent homology can identify qualitative features such as connected components and holes, with some resistance to noise.

### Source excerpt

It's about time we got back to computational topology. Previously in this series we endured a lightning tour of the fundamental group and homology, then we saw how to compute the homology of a simplicial complex using linear algebra. What we really want to do is talk about the inherent shape of data. Homology allows us to compute some qualitative features of a given shape, i.e., find and count the number of connected components or a given shape, or the number of "2-dimensional holes" it has.