Skill UI

Provides expert guidance on designing, implementing, and optimizing cutting-edge computer vision pipelines. Expertise spans real-time object detection (YOLO26), advanced promptable segmentation (SAM 3), and leveraging Vision Language Models (VLMs) for complex visual reasoning. Ideal for building high-performance systems in robotics, autonomous vehicles, and industrial inspection, including 3D reconstruction and edge deployment optimization.

SGLang is a high-performance serving framework for LLMs and VLMs that uses RadixAttention prefix caching to deliver structured outputs (JSON/regex/grammar) and agentic workflows with function calls, providing up to 5× faster inference than vLLM across multi-GPU production deployments.

A comprehensive evaluation suite designed to benchmark the performance of Large Language Models (LLM) and Vision-Language Models (VLM) specifically for home security applications. It covers 143 diverse tests across 16 critical domains, including tool use, prompt injection resistance, contextual reasoning, alert triage, and detailed scene analysis, ensuring robustness for smart home AI systems.

This bridge provides bidirectional integration between Home Assistant (HA) and advanced AI pipelines. It streams HA camera feeds into the AI system for sophisticated analysis (detection, VLM, REID). Conversely, the analysis results—such as detected objects, descriptions, and person identities—are sent back to HA as image_processing entities, enabling powerful automation and smart notifications.

This benchmark evaluates Vision-Language Models (VLMs) on detecting potential safety hazards within indoor environments. It uses static camera frames, simulating real-world fixed security camera monitoring. The test covers 40 scenarios across 5 critical categories, including fire/smoke, electrical risks, trips/falls, child safety, and falling objects, providing a rigorous measure of VLM's practical safety inspection capability.

This comprehensive benchmark suite evaluates the performance of Vision-Language Models (VLM) on complex video anomaly detection tasks using the SmartHome-Bench dataset. It covers a wide range of real-world smart home scenarios, including security breaches, elderly fall detection, and unusual pet behavior. Unlike single-frame analysis, this benchmark requires deep multi-frame video understanding, making it a robust tool for testing the advanced AI capabilities of video monitoring systems.

This skill provides comprehensive guidance on positioning, structuring, and refining computer vision manuscripts for top-tier venues like CVPR. It covers identifying major novelty, required experimental rigor (e.g., ablations, failure modes), framing contributions, and adhering to specific conference submission standards for CV, VLM, and recognition tasks. Use it to refine your submission strategy and ensure your work meets the high bar set by leading AI conferences.

Expert guidance for designing, implementing, and optimizing computer vision pipelines, including YOLO26 detection, SAM 3 segmentation, VLMs, depth estimation, calibration, and edge deployment.

Extract text from scanned PDFs, photographed pages, and images lacking a text layer. Covers OCR backends like Tesseract, PaddleOCR and VLMs, language packs, force-OCR, batch tuning and failure fixes.

Estimate the required memory for inference of Safetensors or GGUF models from the Hugging Face Hub using HTTP Range requests, without downloading weights. Supports optional KV cache estimation for LLMs and VLMs.