Ultralytics YOLO27: Join the waitlist.
Get Started
From a product idea to your first YOLO26 prediction
Foundation pathwaybeginner ~3 hours 10 lessonsFinal exam · Certificate

Computer Vision Foundations

Learn the judgment behind every working vision system — framing the task, choosing among the seven Ultralytics YOLO tasks, auditing data and labels, splitting honestly, and reading metrics — with runnable YOLO26 examples on real images and datasets.

By Ultralytics Academy

Begin courseSign in to save progress
What you'll learn

Frame a computer vision problem, choose the right task and labels, check that your data matches reality, and know what "good" looks like before you train.

  • Turn a product idea into a three-line task spec, and keep business rules in code instead of in the model.

  • Match an output to the right one of seven tasks: classification, detection, instance segmentation, semantic segmentation, depth estimation, pose, or OBB.

  • Read detector output and set confidence and IoU thresholds from the cost of each mistake.

  • Audit a real dataset for coverage gaps, broken labels, and near-duplicate images that leak across splits.

  • Read precision, recall, mAP50-95, per-class AP, and the confusion matrix together to decide ship, iterate, or rethink.

What you'll build

  • A task spec for your project, with a decision rule written as code on top of YOLO26 output.

  • A dataset audit and label check for a real dataset, with a ranked collection plan.

  • A leak check and a group split that keep near-identical images out of validation.

  • An evaluation read-out that names the bottleneck class and the shipping threshold.

  • A first YOLO26 run on your own images, checked against your spec.

Prerequisites

  • No prior computer vision experience — the concepts stand on their own.

  • Optional but recommended: Python 3.10+ and pip install ultralytics to run the examples. Everything runs on a laptop CPU in seconds.

Course content

4 modules · 10 lessons

Module 1

Frame the Problem
1

Frame the Problem

Turn a fuzzy product idea into a three-line task spec, and keep the business rule in code where you can change it without retraining.

Module 2

Pick a Task
2

The Seven Vision Tasks

Map an output line to classification, detection, instance segmentation, semantic segmentation, depth estimation, pose, or OBB, and pick the YOLO26 weights to start with.

3

Understand Object Detection

Read a detector's output — class, box, confidence — and set the confidence and IoU thresholds from the cost of each mistake.

4

Boxes vs Masks: When You Actually Need Segmentation

Measure what a mask adds over a box on a real image, and decide from the downstream consumer whether your project needs shape at all.

5

Pose Estimation and OBB

Read keypoints and rotated boxes from YOLO26, write a rule on top of each, and recognize when posture or angle is the real signal.

Module 3

Datasets and Labels
6

Datasets that Reflect Reality

Audit a real dataset for coverage gaps before you label more, and turn the gaps into a collection plan with hard negatives.

7

Annotation Quality

Know what a YOLO label contains, lint label files for broken rows, and keep annotators consistent with a guideline and a measured calibration round.

Module 4

Read the Metrics
8

Splits that Tell the Truth

Find near-duplicate images that leak across train and validation, and split by the unit that will be new at deployment.

9

Reading Detection Metrics Honestly

Run validation, read precision, recall, mAP50-95, per-class AP, and the confusion matrix together, and pick a shipping threshold from a target.

10

From Lab to Production

Install Ultralytics, run YOLO26 on your own images, check the output against your spec, and choose your next step.