Latest news, developments and reporting on Computer Vision.
An MIT Technology Review investigation says it documented more than 1,000 people who crossed areas watched by US border surveillance towers unapprehended and died there, including within range of new AI-powered towers.
AMD has agreed to acquire Fei-Fei Li's World Labs in an all-stock deal valued at about $8.2 billion, bringing the world-model lab and its founder into the chipmaker as it competes with Nvidia in AI.
TechRadar: Anker unveiled Eufy S4 video doorbell with VR viewing and AI tracking. No pricing, availability, or specs were included.
Robotic dogs are moving from security patrols into fields and greenhouses, where their cameras and onboard AI map crops, spot weeds and flag disease.
TECNO and Angélica Dass have unveiled “100 Portraits of Becoming,” a two-year project across five countries. It begins with 20 Kenyan portraits and will total 100 stories.
Google Research launched a four-framework AI video co-director for coherent, minutes-long multi-shot video.
Engadget's guide explains how iPhone users can use Live Text through the Camera app to copy, share, translate or look up text, and how the iOS 15 feature differs from Visual Intelligence.
TechRadar reviews the Reolink OMVI 3i, a mains-powered camera with 180-degree panoramic views, 4K tracking and edge AI.
Scotland Yard says advances in CCTV enhancement technology have allowed detectives to re-examine 27-year-old footage from Putney Bridge station in the unsolved 1999 murder of BBC presenter Jill Dando, though police say it is not a new investigation.
Aosu's R1 Ultra doorbell offers 60-day local storage and AI detection without a subscription, TechRadar says.
MarkTechPost details an AugLy workflow for multimodal augmentation and adversarial robustness in images, text and audio.
Liquid AI released LFM2.5-VL-3B-DSpark, an experimental speculative-decoding drafter for its LFM2.5-VL-3B vision-language model, claiming up to 3.13x faster decoding on Apple silicon and 2.66x on an H100 with unchanged output.
Apple Intelligence for Home adds AI-generated text descriptions and other features to HomeKit Secure Video. The Verge's test against Ring and Nest found the alerts more useful, but Apple's subscription pricing runs up to three times higher.
Five Indianapolis officers face criminal charges after authorities say they misused Flock license-plate-reading cameras, including to track exes and women. One also faces stalking counts.
A 9to5Mac comparison tested iOS 27’s Clean Up and Extend tools against Adobe Photoshop’s Firefly generative fill and expand on six photos, finding mixed results.
Cal AI co-founder Zach Yadegari, whose photo-based nutrition app passed 15 million downloads and was acquired by MyFitnessPal, will discuss viral growth at TechCrunch Disrupt 2026 in San Francisco.
Liquid AI released LFM2.5-VL-DSpark, a 279.5M-parameter speculative-decoding draft model for its 3B vision-language model, claiming on-device decode speedups of up to 3.13x and H100 gains of up to 2.66x, with day-one support for llama.cpp, MLX-VLM and SGLang.
Revolut pilots Pay with Smile, a UK first in-store facial recognition checkout, at select London coffee shops.
Revolut is preparing a facial recognition payment system for shops and hospitality venues in the UK, letting customers pay by looking at a checkout terminal, Sifted reports.
TechRadar reports experts warning that popular TP-Link home security cameras could be hacked for spying, and that Interpol used new facial recognition tools to identify 126 terrorists from over 100,000 images.
TechRadar reports the Obsbot Meet 2 4K webcam with AI framing is 23% off, at $99 in the U.S. and £99 in the U.K.
AIsphere has released PixVerse R2, a real-time world model that puts video, audio, action and text control into one causal autoregressive framework. The company says general capability and real-time speed can now scale together.
ECCV 2026 opened in Malmö, Sweden, drawing more than 7,000 attendees and over 10,000 submissions. The conference awarded its Koenderink Prize to a Stanford team led by Fei-Fei Li and recognized foundational work in perceptual loss, object detection and continual learning.
In September 2026, LeEnjoy’s live barbecue test, Figure’s Helix 2.5 home trials, human-video pretraining work and Yann LeCun’s ECCV keynote pushed embodied AI toward world models and away from pure action prediction.