YOLO

YOLO (You Only Look Once) is a state-of-the-art, real-time object detection algorithm that can quickly detect and locate objects within an image or video. The YOLO architecture works by taking an input and separating it into a grid of cells and each of these cells is in charge of detecting objects within that region. YOLO returns the bounding boxes containing all the objects in the image and predicts the probability of an object being in each of the boxes and also predicts a class probability to help identify the type of object it is. YOLO is a highly effective object detection algorithm and making YOLO and open-source project led the community to make several improvements in such a limited time.

General
Relese date	2015
Author	Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi
Paper	(https://arxiv.org/abs/1506.02640)
Type	Object detection algorithm

YOLO - Resources

Learn even more about YOLO!

v7 Labs Blog "YOLO: Algorithm for Object Detection Explained".
YOLOv5 Repository Object detection architectures and models pretrained on the COCO dataset.
YOLOv6 Web demo Gradio demo for YOLOv6 for object detection on videos.
Hugging Face Spaces Test YOLOv7 in the browser with Hugging Face Spaces.

Assembly

Plants already collect sensor histories, vibration recordings, camera images, maintenance logs and QC inspections. Turning that into a working model normally takes an ML team weeks. With Assembly, an operator writes something like: > Prevent unplanned stoppages on Press Line 3 and stop defective metal nuts from reaching assembly. > Max 1 false alarm per shift; catch at least 90% of failures and defects. The agent then: 1. Formulates concrete ML tasks with measurable KPIs (e.g. "will this press fail within 30 cycles?", "where is the defect on this part?"). 2. Discovers which plant streams matter for each task. 3. Designs each dataset by writing its own `prepare.py`; validators check it for leakage and sanity before an operator approves it. 4. Researches models in an autoresearch-style loop: it explores the data, plans experiments, writes train.py, runs it under a time budget, and keeps a change only if fixed scoring code, comparing against labels the agent never sees, says it's better. 5. Ships the winner: full training, a final gate on a hidden test split, packaging and a live endpoint.

Lynceus - AI Gate Inspector for RDF Trucks

Cement plants burn refuse-derived fuel (RDF) to replace coal, but every truck of shredded waste can hide metal, stone or tyres. One metal chunk that reaches the shredder can stop the kiln, costing up to Rp 798 M per incident. Today a gate officer checks each load by eye and keeps no record. Lynceus moves that check to the weighbridge, before the load is tipped. The operator uploads the load photo and the waybill. A vision model on AMD Token Factory reads the waybill and auto-fills the register (unreadable fields stay empty, never guessed), then finds and boxes contaminants in the load. A YOLO26m detector on the gate PC’s CPU gives a second opinion. The model observes; the rule engine decides. Six gate checks (visual contaminants, waybill, waybill vs load, gas sensors, weight, detector) produce one verdict that says which check failed and why. Each verdict comes with SOP-cited next steps, one-click supplier notice or escalation, and Pythia, an assistant that answers only from plant documents with citations, or says “cannot answer”. Every inspection is sealed as a tamper-evident SHA-256 audit record. Measured live on 80 labelled RealWaste photos: 19 of 20 metal loads stopped, 35 of 40 clean loads passed, ~20 s median to a verdict. Per 1,000 trucks, hard loads reaching the shredder drop from 6 to 1 and trucks checked by hand from 1,000 to 142. All live AI runs on AMD GPUs via AMD Token Factory (DeepSeek-V4-Flash-Vision-Exp, DeepSeek-V4-Flash), with Google Gemma 4 as the fallback vision model and YOLO26m fine-tuned on Google Colab. Unlike magnets or sorting-line AI, Lynceus acts before tipping. Target market: Indonesia first, then the EU, at €500 per gate per month. Built by Rafi Cahya Mahardika and Muhammad Muqorrobin Al Anshori for the AMD Developer Hackathon ACT III, Intelligent Industry. lynceus.web.id

Crash Guard: AI Traffic Accident Detection System

Crash Guard is an AI-powered traffic monitoring and emergency response system designed to improve road safety by detecting accidents and helping emergency services respond faster. The Camera Panel is built with Python, OpenCV, and a custom fine-tuned YOLOv8 deep learning model that analyzes traffic video feeds to identify vehicle accidents and classify their severity. When an accident is detected, the system captures video footage from before and after the incident, uploads the evidence to Cloudinary, and stores incident information, including severity, confidence score, timestamp, and GPS coordinates, in Firebase Firestore. Firebase Cloud Messaging (FCM) delivers real-time push notifications to the connected Flutter mobile and web Management Panel. This interface enables emergency responders and administrators to review incident reports, watch recorded evidence, locate accidents on maps, and track incident status. The system uses background processing to handle video uploads and cloud operations without interrupting the monitoring interface. By combining computer vision, cloud infrastructure, and cross-platform application development, Crash Guard aims to reduce delays in accident reporting and support faster coordination among emergency services.

Sentry: AI Yield Copilot for Semiconductor Fabs

Semiconductor fabs often find a broken tool weeks after it breaks. An etch chamber's edge ring wears, nothing alarms, and every 25-wafer lot it processes is damaged until wafer sort catches it, often around 12 weeks later. At roughly $20,000 per advanced-node wafer, each of those lots puts about $500,000 at risk. Sentry closes that gap. Every wafer map has a shape, and the shape points to the cause: an edge ring suggests the etch or thermal chamber, a scratch suggests the handling robot or CMP polisher. Sentry classifies each map into one of 9 WM-811K patterns with a calibrated 3-CNN ensemble (0.912 macro-F1, 9.5 ms per wafer on CPU, scored on lots it never trained on). When one tool produces the same pattern three times in 24 hours, it raises an alert showing the evidence wafers, the likely cause with its source, the machine to check and the dollars exposed. Ask Sentry, an open LLM agent served on AMD GPUs through AMD Token Factory, answers questions about the line in plain English using 11 tools over live line data. Uncertain wafers (1.1%) go to a person, and the engineer decides whether to hold lots, open a ticket or dismiss, with every decision logged. In our replay of 500 real wafers through four simulated etch tools, Sentry flagged the drifting tool on its 3rd Edge-Ring wafer with zero false alarms beforehand. Built with PyTorch, FastAPI, Docker and Hugging Face Spaces.

Trivela Labs

Most football teams can’t afford match analysis. The tracking systems cost thousands, and tagging a game by hand takes an analyst hours. Trivela Labs does it from one match video. No trackers, no sensors, no GPS vests, no extra hardware. Upload the footage and our Tactical Engine goes to work. It runs YOLO to detect and track every player, the goalkeepers, the referee and the ball, frame by frame. It splits the teams by kit color and turns all that movement into a full match report with more than 25 stats. For the attack, you get passes, passing accuracy, long balls, crosses, dribbles, shots, shots on target, xG and touches in the opposition box, plus a passing map and a shots map. For the defence, you get tackles, saves, dispossessions and a defensive contribution score. Beyond those are possession, touches, corners, free kicks, time in the final third, distance covered, number of sprints, sprint speed and ball speed. Heat maps show where each player operates, with and without the ball. Coaches and analysts can use it straight away. It works on any match video, from a youth game filmed on a phone to a broadcast feed. The report is saved to the cloud, and the dashboard keeps every past match, player and team in one place, so you can track form over a season. You can also just ask. The AI match assistant is powered by Google Gemma and runs on AMD hardware. Ask “Who dominated midfield?”, “Why did we lose the ball so often in the second half?” or “Who covered the most ground?” and it answers like a coach, using the numbers from your match. It stays on football and sticks to what the engine measured. If we don’t track something, like fouls, it says so and offers the closest stat it does have. Built with YOLO, Claude Code, AMD, Google Argon 4, Google Gemma, GitHub, and blood, sweat, and tears.

Super-Infrared Vultr RAG - 6 Filter Agent

Super-Infrared v3.1: 2-mode + Bio Healthy + Ruang Teduh + SDM - No Phone Has This Spec. Inspired by Pasar Kebayoran Lama fire risk & garment factories Jakarta. SAFETY 4.7s: MLX90640 32x24 thermal (300C surface, 612C WHITE HOT), IMX219 8MP + 850nm IR fusion CCTV, ESP32-S3 relay buzzer 95dB + shutdown. 4.7s response for factory/mall/kos. AIR+BIO HEALTHY: NDIR 4-ch 4.26um CO2, 3.3um H2O, 3.4um Naphtol, 7.7um emissivity + SHT41 Swiss ±0.2C ±1.8% + MQ135 gas. Ideal 22-26C RH40-60% CO2<1000ppm. If >1000ppm: auto fan+relay+WhatsApp. Ruang Teduh 12-City AQI - shaded QC area for garment & obsidian altarum. SDM RECRUITMENT (Vultr Agent Rush HR track - required by jury): 1) Candidate screening via AS7341 spectral color blindness test + HX711 weight 4067/39, 2) Interview scheduling buzzer+Telegram+Meet, 3) Onboarding checklist training device, 4) Workflow tracking IMX219+HX711 KPI. All via super-infrared.local on phone, no extra tools. BRAIN: RPi5 edge AMD ROCm+Ryzen AI (Delegate not just Operate) + Vultr Cloud Inference AMD Developer Cloud EPYC+MI355X+Cisco Silicon One. Pattern: TensorWave 2x perf 40-60% saving, AT&T single-GPU millions, Cisco Healthcare. Open by Design, no CUDA lock-in, More AI/Watt/Dollar. Proof: amd_check.py + ROCm 6GW validation in repo. Repo super-secure-syst-v23 32 commits: gallery, Bio_Healthy_Room_Spec.md, Ruang_Teduh_Spec.md, SDM_Rekrutmen_Module.md. Pricing v3.0 outright system & technique license Rp1.5-3.5M (not hardware MLX Rp1.07M). Demo super-infrared.local, OTA, MQTT, Tailscale, Telegram /status. No phone has NDIR+MQ135+SHT41+MLX+AS7341+HX711.