# Ultralytics Academy

> Free computer vision and Ultralytics YOLO courses with hands-on lessons, projects, and certificates.

## Computer Vision Foundations
- [Frame the Problem](https://academy.ultralytics.com/courses/computer-vision-foundations/frame-the-problem): Turn a fuzzy product idea into a single sentence the model can answer.
- [The Six Vision Tasks](https://academy.ultralytics.com/courses/computer-vision-foundations/vision-tasks): Match an output sentence to classification, detection, instance segmentation, semantic segmentation, pose, or oriented bounding boxes.
- [Understand Object Detection](https://academy.ultralytics.com/courses/computer-vision-foundations/object-detection): Boxes, classes, and confidence — what detection actually returns and what it does not.
- [Boxes vs Masks: When You Actually Need Segmentation](https://academy.ultralytics.com/courses/computer-vision-foundations/instance-segmentation): Move from rectangles to pixels — and only when the shape genuinely matters.
- [Pose Estimation and OBB](https://academy.ultralytics.com/courses/computer-vision-foundations/pose-and-obb): Two specialized tasks and the kinds of problems where they shine.
- [Datasets that Reflect Reality](https://academy.ultralytics.com/courses/computer-vision-foundations/datasets-that-reflect-reality): The model only knows what's in the dataset. Build the dataset like the model depends on it.
- [Annotation Quality](https://academy.ultralytics.com/courses/computer-vision-foundations/annotation-quality): Half a million boxes are worse than fifty thousand consistent ones.
- [Splits that Tell the Truth](https://academy.ultralytics.com/courses/computer-vision-foundations/splits-that-tell-the-truth): Why training, validation, and test splits exist — and the leak that ruins half of them.
- [Reading Detection Metrics Honestly](https://academy.ultralytics.com/courses/computer-vision-foundations/reading-detection-metrics): Precision, recall, mAP — what they hide and how to look at them together.
- [From Lab to Production](https://academy.ultralytics.com/courses/computer-vision-foundations/from-lab-to-production): The smallest end-to-end loop: install, predict, measure, decide.

## Building High-Performance YOLO Datasets
- [Start With the Business Objective](https://academy.ultralytics.com/courses/dataset-readiness-for-yolo/start-with-the-business-objective): Translate a business goal into a vision task, a class list, and a success metric before any image is collected.
- [Define the Dataset Specification](https://academy.ultralytics.com/courses/dataset-readiness-for-yolo/define-the-dataset-spec): Plan the dataset on paper before a single image is collected — scenarios, environments, edge cases, and negatives.
- [Collect High-Quality, Representative Data](https://academy.ultralytics.com/courses/dataset-readiness-for-yolo/collect-representative-data): Capture real production conditions, not staged-only images — and avoid the collection biases that quietly kill models.
- [Annotation Best Practices](https://academy.ultralytics.com/courses/dataset-readiness-for-yolo/annotation-best-practices): Write a labeling guide, pick a tool, and apply rules that keep 10 annotators producing the same labels.
- [Dataset Quality Control](https://academy.ultralytics.com/courses/dataset-readiness-for-yolo/dataset-quality-control): Find missing labels, wrong classes, sloppy boxes, duplicates, and class imbalance — before you train.
- [Split the Dataset Correctly](https://academy.ultralytics.com/courses/dataset-readiness-for-yolo/split-the-dataset-correctly): Train, val, and test — and the leakage rules that decide whether your metrics are honest.
- [Use Augmentation Carefully](https://academy.ultralytics.com/courses/dataset-readiness-for-yolo/use-augmentation-carefully): Augmentation extends variance — it doesn't replace real data. The rules for picking transforms that match deployment conditions.
- [Know When the Dataset Is Ready](https://academy.ultralytics.com/courses/dataset-readiness-for-yolo/dataset-readiness-checklist): The single checklist that decides whether to train now or fix the dataset first.
- [First Fine-Tune and the Iteration Loop](https://academy.ultralytics.com/courses/dataset-readiness-for-yolo/first-fine-tune-and-iterate): Use pretrained YOLO26 weights, train with defaults, and let the validation results tell you what to fix in the dataset.
- [Enterprise Client Checklist](https://academy.ultralytics.com/courses/dataset-readiness-for-yolo/enterprise-client-checklist): A reusable, customer-facing checklist that CS, Platform, Docs, and Academy can all point to.

## Train your first YOLO model
- [Install Ultralytics](https://academy.ultralytics.com/courses/train-your-first-yolo/install-ultralytics): Get a working environment in one command — and verify GPU access.
- [Predict on an Image](https://academy.ultralytics.com/courses/train-your-first-yolo/predict-on-image): Run a pretrained Ultralytics YOLO model and read the result object — boxes, confidences, classes.
- [Choose an Ultralytics YOLO Model Size](https://academy.ultralytics.com/courses/train-your-first-yolo/choose-a-model-size): Nano, small, medium, large, x-large — what each one costs and where each one wins.
- [Prepare a Custom Dataset](https://academy.ultralytics.com/courses/train-your-first-yolo/prepare-a-dataset): Convert images and labels into the Ultralytics YOLO format — folders, files, normalized coordinates.
- [Write a Dataset YAML](https://academy.ultralytics.com/courses/train-your-first-yolo/write-the-data-yaml): The single config file that connects your folders to YOLO's training loop.
- [Train Your First Model](https://academy.ultralytics.com/courses/train-your-first-yolo/train-your-first-model): One Python call, sensible defaults, and a checkpoint at the end.
- [Validate and Diagnose](https://academy.ultralytics.com/courses/train-your-first-yolo/validate-and-diagnose): Beyond the single mAP number — per-class AP, confusion matrices, and what they tell you to do next.
- [Resume and Iterate](https://academy.ultralytics.com/courses/train-your-first-yolo/resume-and-iterate): Continue from where training stopped, fine-tune best.pt, and run experiments cheaply.
- [Run Inference on Video](https://academy.ultralytics.com/courses/train-your-first-yolo/video-inference): From single-image predict to real-time video — and what changes when frames are continuous.
- [Export for Deployment](https://academy.ultralytics.com/courses/train-your-first-yolo/export-for-deployment): ONNX, TensorRT, CoreML, OpenVINO, QNN — pick the format your runtime actually wants.

## Ultralytics YOLO in Production
- [Choose a Deployment Target](https://academy.ultralytics.com/courses/yolo-in-production/choose-a-deployment-target): Map your hardware and constraints to the runtime that wins on it.
- [Export to ONNX (and Verify Parity)](https://academy.ultralytics.com/courses/yolo-in-production/export-to-onnx): The portable starting point — and how to confirm the export didn't quietly break the model.
- [Optimize with TensorRT](https://academy.ultralytics.com/courses/yolo-in-production/optimize-with-tensorrt): FP16 and INT8 — when each one wins, and what to test before you ship.
- [OpenVINO on CPU](https://academy.ultralytics.com/courses/yolo-in-production/openvino-on-cpu): Real CPU speedups via Intel's optimized kernels — useful when GPUs aren't an option.
- [Object Tracking with ByteTrack and BoT-SORT](https://academy.ultralytics.com/courses/yolo-in-production/tracking-with-bytetrack-and-botsort): Persistent IDs across frames — the foundation of every counting, alerting, and analytics pipeline.
- [Counting, Heatmaps, and Speed Estimation](https://academy.ultralytics.com/courses/yolo-in-production/counting-heatmaps-and-speed): Three of the most common downstream pipelines — and the geometry behind them.
- [Multi-Stream Inference](https://academy.ultralytics.com/courses/yolo-in-production/multi-stream-inference): Run multiple cameras concurrently without dropping frames.
- [Observability and Drift](https://academy.ultralytics.com/courses/yolo-in-production/observability-and-drift): What to log, what to dashboard, and how to spot accuracy regression before users do.
- [Cost and Latency Tuning](https://academy.ultralytics.com/courses/yolo-in-production/cost-and-latency-tuning): The handful of knobs that move the needle, and the ones that don't.

## Build with Ultralytics Platform
- [Why Ultralytics Platform](https://academy.ultralytics.com/courses/build-with-ultralytics-platform/why-ultralytics-platform): What Platform is, what it isn't, and the lifecycle it covers.
- [Add and Prepare Data](https://academy.ultralytics.com/courses/build-with-ultralytics-platform/upload-and-prepare-data): Upload data or index connected cloud storage in place, then prepare a Platform dataset.
- [Smart Annotation as an Accelerator](https://academy.ultralytics.com/courses/build-with-ultralytics-platform/smart-annotation): Let SAM (or your fine-tuned YOLO) propose annotations, let humans review — and the rules that keep this from poisoning your dataset.
- [Curate, Dedup, and Split](https://academy.ultralytics.com/courses/build-with-ultralytics-platform/curate-and-review): From a labeled pile to a clean train/val/test ready for training.
- [Cloud Training](https://academy.ultralytics.com/courses/build-with-ultralytics-platform/cloud-training): One click, one model size, one GPU class — and a trained model when you come back.
- [Track Experiments](https://academy.ultralytics.com/courses/build-with-ultralytics-platform/track-experiments): Compare runs, name them well, and keep the lessons.
- [Deploy a Model](https://academy.ultralytics.com/courses/build-with-ultralytics-platform/deploy-a-model): From best.pt to a managed inference endpoint with one click.
- [Monitor in Production](https://academy.ultralytics.com/courses/build-with-ultralytics-platform/monitor-in-production): Watch latency, detection volume, and drift — and know when to retrain.
- [Privacy, Regions, and Compliance](https://academy.ultralytics.com/courses/build-with-ultralytics-platform/regions-and-compliance): Where your data lives, where compute happens, and what dataRegion means in practice.
- [From Prototype to Pipeline](https://academy.ultralytics.com/courses/build-with-ultralytics-platform/from-prototype-to-pipeline): The repeating shape of a CV team's quarter — and the artifacts that make it sustainable.

## Optional
- [Computer Vision Foundations](https://academy.ultralytics.com/courses/computer-vision-foundations): From the first image to the first model
- [Building High-Performance YOLO Datasets](https://academy.ultralytics.com/courses/dataset-readiness-for-yolo): From problem definition to training readiness
- [Train your first YOLO model](https://academy.ultralytics.com/courses/train-your-first-yolo): From pip install to a custom-trained detector
- [Ultralytics YOLO in Production](https://academy.ultralytics.com/courses/yolo-in-production): From best.pt to a real, observable, optimized system
- [Build with Ultralytics Platform](https://academy.ultralytics.com/courses/build-with-ultralytics-platform): End-to-end CV — from raw frames to a deployed, monitored model