Tag
A look at Labelme, an open-source Python/Qt image annotation tool for drawing polygons, boxes, circles, and points, featuring AI-assisted annotation via SAM, EfficientSAM, YOLO-World, and SAM3, plus VOC/CCO-format exports and video annotation support.
A curated list of 10 open-source GitHub repositories that use AI to extract insights from satellite imagery, covering tasks like segmentation, flood mapping, tree detection, and land-cover analysis for Earth observation.
The paper introduces a controlled evaluation framework separating unique class support from representation allocation to determine when encoder adaptation remains worthwhile after classifier correction in long-tailed recognition, showing benchmark construction alone can flip this training decision.
Vathography is a new creative approach that turns a photograph's inferred depth map into an artwork, using a diffusion transformer to estimate depth and letting authors treat focus as a mapping of depth to tone and color. Inspired by Marigold V2's detailed depth decomposition, it reframes a normally hidden technical pass as a form of artistic expression.
video-grep is a terminal-first tool that lets you search video content by describing actions or scenes in plain English, indexing frames locally to find exact timestamps without leaving the SSH session.
论文提出 Bottle-SPU（Bottle-SKU）大规模多SKU洗发水瓶表面缺陷检测基准，包含13,119张图像与20,419个实例级标注框，并提出原型引导的动态域调制方法 ProDMM，通过SKU视觉原型生成低秩Query/Value调制以提升跨产品变体的缺陷检测泛化能力。
Shire Intelligence turns existing restaurant CCTV cameras into a real-time operations layer, tracking table states, server workloads, and guest needs to cut table turn times by 12% and save restaurants ~$31,000 annually. The Y Combinator-backed startup is now live in 13 restaurants and expanding its AI-driven POS, scheduling, and reservation platform.
ReLift reframes dense 3D tracking from monocular video as geometric trajectory repair, using a feed-forward trajectory corrector that alternates attention along and across lifted trajectories. It achieves state-of-the-art mean accuracy on WorldTrack, improving over the strongest baselines by 2.1 APD on all points and 3.0 on dynamic points without image-based training.
Arrowmax's Arrowspace S1 electric screwdriver uses a built-in camera and AI to identify screw type and size, recommend the right bit, and auto-select a safe torque setting, launching via Kickstarter ahead of an April 2027 ship date.
University of Rochester and NVIDIA researchers developed a pixel-space diffusion model enabling precise, localized edits to AI-generated images without unintentionally altering unrelated details, unlike latent-space models used by systems like ChatGPT and Google Gemini.
A dating app called Rivet pairs users and shows the match to other users who vote yes or no, using an AI-generated blended image of the two people as a 'couple photo' to drive the verdict, raising questions about whether surfacing appearance bias openly is better than hiding it in ranking algorithms.
The paper proposes FDSML, a formula-driven supervised metric learning approach that uses parameter-aware triplet loss on a refined fractal dataset (FineFractalDB) to improve synthetic pre-training for image retrieval, achieving notable Recall@1 gains over FDSL on Stanford Online Products and Stanford Cars.
AB-Attention提出一种免训练的块稀疏注意力方法，通过块内聚类选择、包含概率加权采样和细粒度核收集来近似稠密注意力，在Wan2.1和HunyuanVideo上跳过89-92%注意力计算，同时PSNR提升最高4.4dB，端到端加速最高2.23倍。
The paper introduces ImagiRep, a simulator-grounded benchmark showing that spatial reasoning benefits from diverse representations beyond RGB, and proposes GEARS, a lightweight test-time strategy for adaptively selecting representations to improve reasoning about unobserved states.
Sperid Labs 发布 Idol，能将单张物体照片在 2 秒内转换为包含被遮挡面的 3D Gaussian splat（最多约 100 万个高斯），可在消费级 GPU 上以 7.2GB 显存运行。
提出了 Augmentation-Guided Adversarial Training (AGAT) 框架，利用基于增强的局部稳定性 (ALS) 得分选择不稳定邻域的样本进行对抗训练，在仅对 20–30% 样本施加对抗监督的情况下达到相当或更高的鲁棒准确率，同时降低训练成本。
A successor to the Tesseract OCR engine has been released as indra by datalab-to, notable for being just 1M parameters while showing strong gains in both quality and speed on benchmarks.
Hi3D released Hi3D V1.0, an open-source 3D model built on Sparc3D that adds detail to rough 3D meshes from a reference image, with inference code and geometry refinement weights available on GitHub, while texture generation remains in the commercial V3.
Unlimited OCR 2 is currently the top trending paper on Papers with Code, with a chat interface available for the paper's PDF.
论文提出 Isharah-Selfie 数据集，这是一个完全通过前置智能手机自拍视角采集的大型阿拉伯手语（ArSL）数据集，包含 16 名签名者的 8,997 段视频和 1,493 个句子，用于连续手语识别与翻译，并揭示现有方法在自拍条件及未见句子泛化上的不足。