YOLOv8

Top Builders

Explore the top contributors showcasing the highest number of app submissions within our community.

YOLO v8

Ultralytics YOLOv8 is a cutting-edge, state-of-the-art (SOTA) model that builds upon the success of previous YOLO versions and introduces new features and improvements to further boost performance and flexibility. YOLOv8 is designed to be fast, accurate, and easy to use, making it an excellent choice for a wide range of object detection and tracking, instance segmentation, image classification and pose estimation tasks.

What's new in YOLOv8?

YOLOv8 supports a full range of vision AI tasks, including detection, segmentation, pose estimation, tracking, and classification. This versatility allows users to leverage YOLOv8's capabilities across diverse applications and domains.

General
Relese date	May, 2023
Repository	https://github.com/ultralytics/ultralytics
Type	Real time object detection

Libraries

Discover YOLOv8

Documentation YOLOv8:
Quickstart Quickstart with YOLOv8
GitHub Repository View the GitHub repository for YOLOv8
Hugging Face Spaces Test YOLOv8 in the browser with Hugging Face Spaces

Edit on GitHub

YOLO YOLOv8 AI technology Hackathon projects

Discover innovative solutions crafted with YOLO YOLOv8 AI technology, developed by our community members during our engaging hackathons.

TempoGraph: Local Multimodal Video Analysis

TempoGraph is a fully-local, privacy-preserving multimodal video analysis system that turns raw video files into rich structured outputs — entities, behaviors, transcripts, timelines, and interactive knowledge graphs — without sending a single frame to the cloud. Stage 1 — Frame Selection: Motion-aware sampling with static, moving, and auto camera modes. For moving cameras it estimates homography to separate object motion from camera movement, then identifies keyframes where motion peaks exceed a configurable sigma threshold. Stage 1.5 — Audio Transcription: Whisper.cpp running on Vulkan transcribes the full audio track to millisecond-accurate segments. Stage 2 — YOLO Detection: YOLO26 runs on 2nd GPU over every sampled frame, outputting normalized bounding boxes, class names, track IDs, and confidence scores. Stage 3 — Depth Estimation: Depth Anything V2 via HuggingFace Transformers adds per-detection mean depth to every bounding box, giving 3D spatial context to 2D detections. Stage 4 — Frame Scoring: Picks which frames the VLM actually sees. In keyframes mode, only motion-peak frames are forwarded. In scored mode, FrameScorer ranks all YOLO-scanned frames using a weighted combination of motion delta, new YOLO class appearances, tracked object churn, and IoU drop between frames — then fills the VLM budget with the highest-signal frames. Keyframes are always pinned in first regardless of mode. Stage 5 — VLM Captioning: Qwen3.5-VL-9B served by a custom llama.cpp build compiled for AMD ROCm/HIP, running on an AMD RX 9070 XT with a 100k-token context window. Frames are chunked and sent to the model alongside YOLO-derived annotations. Each chunk's summary seeds the next prompt for narrative continuity across the video. Stage 6 — Aggregation: A final text-only LLM call synthesizes all per-chunk captions and the audio transcript into a structured JSON with entities, visual events, audio events, and multimodal correlations linking what was said to what was seen.

AssemblyMind

Assembly Mind is a real-time multimodal agent for hardware assembly auditing. In electronics prototyping, simple assembly errors, reversed polarity, incorrect connections, component mismatches, are a leading cause of board failure. These errors are typically caught only after power-on, resulting in destroyed components and hours of debugging. Existing Automated Optical Inspection systems are designed for high-volume manufacturing floors and require dedicated fixtures, pre-programmed rules, and substantial capital investment. They do not serve the engineer working on a one-off prototype at a workbench. Assembly Mind addresses this gap by combining schematic document understanding with live visual perception. The system ingests schematic PDFs or netlists and processes live camera feeds from a standard webcam. A vision-language model, running on AMD Instinct MI300X GPUs via ROCm, performs semantic reasoning across the schematic and physical assembly to detect discrepancies before power is applied. The agent outputs structured audit results and natural-language guidance, flagging errors such as incorrect component orientation or mismatched pin connections. For this hackathon, we demonstrate an end-to-end pipeline: schematic ingestion, real-time camera analysis, multimodal reasoning, and immediate feedback. Built with Qwen2.5-VL, LangChain for agentic orchestration, and Hugging Face Optimum-AMD for ROCm optimization, the system utilizes the MI300X's high memory bandwidth and large HBM3 capacity for efficient high-resolution image processing and long-context multimodal inference. The result is a practical tool that reduces rework time and prevents costly prototyping failures.

CUDA→ROCm Migration Agent

rocm-migrate is a comprehensive toolchain designed to accelerate the transition from NVIDIA CUDA to AMD ROCm environments. At its core, it features a sophisticated two-phase migration agent: a high-speed, rule-based pass handles deterministic API replacements, while a multi-agent LLM loop—utilizing the Qwen-based DeepSeek-R1 (DeepSeek-R1-Distill-Qwen-32B) for complex reasoning and Mistral Codestral for execution—resolves intricate code patterns like custom kernels, NVTX profiling, and mixed-precision idioms. The project demonstrates the real-world utility of ROCm through two additional high-performance tracks optimized for AMD MI300X hardware. Track 2 introduces a DINOv2-Large model fine-tuned on the CCMT plant disease dataset, achieving an impressive 97.06% accuracy and 0.9713 F1 score. Track 3 extends this into conversational AI with a Llama 3.2 11B Vision LoRA adapter, providing multimodal diagnostics and treatment guidance from leaf imagery. Together, these tracks showcase a complete ecosystem—from automated infrastructure migration to production-ready computer vision and generative AI applications—validated on the latest AMD hardware and accessible via a live interactive demo on Hugging Face.

Lucen Robotics

We are building a simulation-first autonomous retail robotics system for night-shift restocking and inventory management. Using NVIDIA Isaac Sim, we model realistic supermarket environments where robots detect out-of-stock shelves, navigate store aisles, and execute pick-and-place restocking tasks autonomously. Our perception pipeline combines computer vision models trained on real supermarket data — for out-of-stock detection and 84.25% SKU classification across 207 products — with depth sensing and SLAM-based navigation within the simulated environment. Robots interact with shelf objects to perform concrete restocking tasks including picking products from backroom inventory, transporting them to target aisles, and placing items in correct shelf positions. The system demonstrates repeatable task execution under varying store layouts and stock conditions, with clear performance metrics and basic failure recovery. A web-based operator dashboard provides real-time monitoring, inventory alerts, and analytics, delivering a complete product experience for retail operators.

NeuroPilot

People who could benefit from hands-free control, for example those with limited mobility, often cannot access it. NeuroPilot changes that: open-source BCI middleware that turns a Muse 2 headband into a real control interface. Muse 2 is already connected to the app, and we have a working integration with DJI Tello. Users train in a 3D simulation and in a training environment; when their brainwave pattern matches a chosen action (move left, right, up, down), the backend fires configurable webhooks or WebSockets. We accumulate 5 actions from the headband; YOLOv8 runs on the Tello camera feed to detect obstacles and Tello state. That batch and the vision data are sent to Gemini, which decides the movement. Every command sent to the DJI Tello is the result of an AI API call, so BCI intent is refined by vision and reasoning before it reaches the drone. The stack is software-only and simulation-first: a Next.js frontend, a FastAPI + WebSocket backend, a web Lab (Three.js), a machine/Connect dashboard for status and live feedback, and per-control webhook URLs. No lab hardware or locked-in vendors. Teams can integrate with existing sims, digital twins, or robots via HTTP.

RIFFAI

RIFFAI’s hackathon prototype demonstrates an end-to-end pipeline for energy site intelligence. The prototype ingests multi-source satellite imagery (Google Earth Engine + direct providers), weather feeds and customer training sets; applies an ultra-fast image/geo-metadata alignment pipeline; runs geospatial segmentation (U-Net/CNN) and time-series forecasting (LSTM); and produces explainable suitability scores and a ranked list of candidate polygons for renewable energy deployment. A minimal React web UI or API visualizes results and emits a sample monitoring alert (thermal anomaly or vegetation/water change). The goal is to show that fast metadata processing + energy-tuned models produce practical, developer-ready outputs in near-real time. We’ll provide domain guidance, sample tiles and a small labeled set for teams that request it.