I'm more curious about the hardware acceleration on Snapdragon and ARM. That matters way more for real deployment than any new DNN feature.

Great to hear.

OpenCV was so easy and smooth to set up for doing tasks like generating thumbnails from uploads from arbitrary photo uploads regardless of format (including funky new formats like webp, avif, or heic).

I guess i'm the only one blown away by this announcement and super excited to get back into image processing.

I remember trying to do photo stitching myself (panoramas) then I failed miserably but it's built into opencv ha. I've used quite a bit of OpenCV features eg. laplace variance for an automatic zoom/focusing mechanical lens camera system (steppers) and contour/blob finding for crude color segmentation.

> LLMs and VLMs, Running Inside OpenCV…Qwen 2.5, Gemma 3, PaliGemma, and the GPT-2 / GPT-4 family

Why these specific models / versions?

Same on mobile. I use Apple's Vision framework on-device to find people in photos for a printing app. Sending users' personal photos to an image-model API is a non-starter on privacy, latancy, and per-photo cost alone. Less flexible than a V-LLM, but for "find the people, give me box" it's instant, free, and works offline.

Curious how do people usually use OpenCV with CCTV? (Use cases)

Can it detect the speed of the car without any hand-made measurement ?

> This is not just another incremental release. OpenCV 5 is a major step forward.

Am I the only one that finds this sentence very cheesey?

To the question about OpenCV peformance to measure JPEG image decoding sequentially and as a part of the PyTorch Dataloader.

TL;DR

OpenCV is fast, but torchvision is faster.

https://arxiv.org/abs/2605.08731