AI Briefing#computer-vision
The latest AI and developer news about #computer-vision, with the original source and a short summary.
The latest AI and developer news about #computer-vision, with the original source and a short summary.
We propose a general-purpose model that accurately identifies and counts objects across diverse domains via text queries.
PaddleOCR has unveiled PP-OCRv6, a new OCR model series with significantly improved accuracy and speed.
Kyung Nong built a visual search system that finds pesticide product information from just a photo, using Vision LLM and Amazon OpenSearch.
This research dramatically increases inference speed by dynamically allocating tokens leveraging the temporal redundancy of video.
SCAIL-2, a model that animates characters via video without intermediate representations, has been released.
This covers the historical evolution of facial recognition technology, which began with manual coordinate recording and advanced through automation and statistical analysis.
A research study introduces an efficient Diffusion model that generates high-quality images reflecting the structure of a single image, without requiring any separate training.
OpenCV 5 has been released, significantly modernizing the DNN engine and greatly strengthening ONNX support.
Qwen-Image-Flash demonstrates the importance of optimizing data composition and training pipelines, beyond objective functions, for accelerating image generation models.
An on-device facial recognition model and database structure have been discovered in Meta's app for smart glasses.
We propose a stochastic Gaussian splatting technique that can render hundreds of millions of Gaussians in real time.
A native app has launched that converts photos into 3D Gaussian Splats using the Mac GPU with no separate setup required.
PapersWithCode has added a feature that lets users browse major AI conference papers by task and check related resources.
Nvidia has unveiled 'LocateAnything,' a Vision-Language Grounding model that is 10x faster than Qwen3-VL through parallel box decoding.
Apple will participate in CVPR 2026 to present its latest research, including generative AI and multimodal LLMs.
NVIDIA has unveiled LocateAnything, which applies Parallel Box Decoding (PBD) technology to revolutionize the grounding speed and accuracy of VLMs.
NeuroFlow is proposed to maximize the inference speed and efficiency of Vision Transformers by using EMA to remove redundant tokens in video.
By comparing visual cortex data from humans and macaques, the study analyzed the alignment performance of various learning rules.
BusPatrol is pushing to let police use license plate data collected by school bus AI cameras.
MUSINSA used **Geometric Prior** instead of a VLM to boost clothing detail shot generation speed by **25x**.
Starbucks scrapped its automated inventory management AI program just nine months after launch due to the limitations of computer vision technology.
ByteDance has unveiled Lance, a 3B-scale unified multimodal model that supports understanding, generation, and editing of images and videos.
Introducing PICO, which is optimized for the human visual system and dramatically improves compression ratio and speed compared to existing codecs.
This analyzes the architecture of WordDetectorNet, which uses per-pixel distance regression and DBSCAN instead of an anchor-based approach.