Voxel has integrated Ventral AI, a computer vision engineering firm, into its team after a year-long strategic partnership. Co-founded by Matko Kulaš and Luka Žagar, Ventral brings expertise in real-time vision and vision-language models for moving equipment, transport fleets, data center construction, and heavy industry. The move strengthens Voxel’s engineering capabilities and European expansion. [link]

Arduino has introduced a standalone USB Camera designed for plug-and-play edge AI with its UNO Q and VENTUNO Q computers. It combines HD imaging, an integrated microphone, adjustable ring lighting, and flexible mounting. Developers can build local applications for object detection, gesture recognition, vision-language models, and multimodal AI assistants without complex camera setup. [link]

AppWork has acquired Paraspot AI, a video-based property inspection platform using computer vision and voice intelligence to detect damage, compare move-in and move-out conditions, and generate timestamped reports with estimated repair costs. AppWork plans to integrate the technology into its maintenance platform, helping property managers streamline inspections, documentation, and compliance workflows. [link]

SEA.AI has integrated its maritime machine vision technology into MARTAC’s unmanned surface vessels. Combining electro-optical and thermal imaging with onboard AI inference, the system classifies vessels and other objects in real time, reporting their type, heading, and proximity. The integration supports intelligence, surveillance, security, and patrol missions without continuous human monitoring of video feeds. [link]

TECNALIA and AUTIS have developed an AI-powered computer vision solution that automatically generates inspection masks for automotive paint-defect detection. Tested using more than 650 images spanning six manufacturers, 20 vehicle models, eight plants, and 13 production lines, the approach reduces manual setup effort and helps inspection systems adapt more efficiently across manufacturing environments. [link]

UK-based Alchera Technologies has raised £2.7 million in seed funding to accelerate its AI-powered transport analytics platform. The company helps local authorities and highway operators make better decisions about road networks and public transport. The investment, led by Maven Capital Partners through the British Business Bank’s South West Investment Fund, will support product development and faster customer deployment. [link]

EyePop.ai has partnered with Secure Passage to integrate customizable computer vision into Truman, its identity-first Physical AI operating system. The integration combines video detections with identity, access permissions, schedules, and location data to reduce false alerts and enable contextual responses. Applications include perimeter monitoring, PPE compliance, and tailgating detection, with automated workflows that can notify personnel or dispatch robots and drones. [link]

Bengaluru-based Curium Life has been named to YourStory’s Tech30 2026 cohort. Its first product, SurgiMeasure, uses computer vision, deep learning, anatomical segmentation, and depth estimation to provide real-time anatomical measurements during laparoscopic and robotic surgeries. Tested in eight live surgeries, the system reports 98% instrument-segmentation accuracy and 30 FPS inference. The startup aims to build a broader AI platform for surgical guidance. [link]

Israeli agtech startup BloomX has raised $13 million in Series A funding led by Elah Fund, bringing its total funding to $20 million. Its biomimetic pollination systems combine robotics and AI to improve crop pollination. Robee uses controlled vibrations for blueberries, while Yahav 2400 uses electric fields for avocado orchards. BloomX reports average yield increases of around 20% among customers and plans to expand into the US and additional markets. [link]

Research highlighted by Pork Business explores how computer vision can continuously monitor pigs’ posture, feeding, and activity to help farmers identify behavioural changes earlier. However, inconsistent lighting, shadows, dirty lenses, dust, humidity, and ammonia can affect model reliability. Researchers recommend measuring illumination at pig height, maintaining predictable light-dark cycles, and designing barn-ready systems that deliver actionable alerts. Lighting alone cannot prevent tail-biting; AI must complement human animal-care practices. [link]

Researchers developed a drone-tracking system combining YOLOv8 object detection, an LSTM network for predicting target movement, and Q-learning-optimized visual reasoning. By anticipating target motion rather than reacting to delayed images, the system achieved a 99.02% obstacle-avoidance rate and lower trajectory errors in simulation and real-world UAV tests, with potential applications in power-line inspection and autonomous navigation. [link]

In a recent interview to Talk GovTech, Lumana’s Asaf Dan explains how AI can transform existing municipal surveillance cameras into real-time intelligence systems. On-premises processing reduces cloud and bandwidth costs, improves resilience during internet outages, and supports faster emergency alerts while keeping sensitive footage local and preserving human oversight.

A latest Reddit post discusses about the industries actually need high FPS (>100) object detection

Another reddit post discusses on working as a freelance CV engineer.