What is computer vision? Turning images into useful information
Computer vision is the field of extracting information from images and video. A system might label a photograph, locate objects, identify regions or follow movement. The first decision is what the output should mean, because different questions require different labels and tests.
SchoolWhool's current learning path covers Python, Data Tools, Maths, Statistics and classical Machine Learning. Open lessons' parts and key terms are free to read.
Classification, detection and segmentation
What does an image look like to a program?
A digital image is represented by an array of pixel values. Colour images commonly have several channels; a batch adds another dimension. Before training anything, inspect the array shape, data type, value range and channel order. A picture that displays correctly is not proof that a model receives the representation you intended.
In a parcel exercise, decide whether orientation is part of the label. Rotating an image may be harmless for some tasks and destructive for others. Write down what a valid transformation means before using augmentation.
Computer vision versus image processing
Image processing changes or measures an image through operations such as resizing, filtering, thresholding and edge extraction. Computer vision asks for information about the scene. A vision pipeline may use those operations without a learned model, or combine them with machine learning.
For the parcel exercise, try a controlled background and a simple threshold as a baseline. Then deliberately change the background. This gives you a concrete failure to investigate and a reason to consider a learned representation, rather than assuming that every image task needs a large network.
Plan a first computer vision project
Choose a narrow question
Start with one output, such as sealed versus open. Define ambiguous cases and avoid mixing classification, detection and tracking in a first experiment.
Check data rights and labels
Use images you may use for the exercise. Keep a short annotation guide and review unclear examples instead of silently assigning a convenient label.
Split by the source of similarity
Keep photographs of the same parcel or capture session together. Otherwise nearly identical pictures can appear in both training and evaluation.
Inspect mistakes under changed conditions
Test different backgrounds, lighting and viewpoints. Keep examples of false positives and false negatives, with an explanation of the conditions behind them.
How to evaluate more than a convincing demo
Choose a measure that matches the output. A classification project can inspect a confusion matrix; a detection project must also define when a predicted box counts as a match; segmentation must compare predicted regions with reference masks. Report the evaluation rule alongside the result.
Show ordinary failures in the project write-up. A montage of selected correct predictions does not tell a reader how the system performs across the test set. Explain which capture conditions you evaluated and which remain outside the exercise.
What should you learn before computer vision?
Begin with Python functions and file handling, then numerical arrays and plotting. Add basic linear algebra and the statistics needed to compare predictions fairly. Classical machine learning gives you a way to practise the whole dataset-to-evaluation workflow before studying deep vision architectures.
SchoolWhool's Python, Data Tools, Maths, Statistics and Machine Learning tracks support those prerequisites. A dedicated Computer Vision track is not published; the links here lead to the foundations you can study now.
Where this guide takes you
- Use the examples to understand the task and its evaluation. They are suggested learning exercises, not measured results or production-ready systems.
- The linked SchoolWhool courses teach the foundations. Specialist deep learning, computer vision, NLP, generative AI and reinforcement learning tracks are not published.
- Open lessons' parts and key terms are free to read. A free account opens quizzes and saved progress, plus practice and other material in the free sample modules.
Questions
Is computer vision part of AI?
Yes. Computer vision is an AI problem area, but its methods include geometric and image-processing techniques as well as machine learning and deep learning.
What is the difference between computer vision and deep learning?
Computer vision names a field of problems involving images and video. Deep learning names a modelling approach that can be used for vision, language and other inputs.
Is OpenCV a deep learning model?
No. OpenCV is a computer vision library with image-processing, geometry and other tools, including facilities for running neural networks.
Should I start with classification or object detection?
Classification usually offers a simpler first exercise because you need image-level labels rather than object locations. Choose detection when locating objects is part of the question.
Can I learn computer vision with Python?
Yes. Python lets you practise loading images, inspecting arrays and evaluating models. Learn those foundations before making a large architecture your first project.
Sources
Technical references and further reading from the authors, educators and maintainers below. Read dates record when these references were checked; the examples and suggested exercises in this guide are SchoolWhool's own.
- Stanford CS231n: image classification and validationRead 2026-10-14
- OpenCV: official tutorialsRead 2026-10-14
Related reading
Deep learning explainedData Tools: arrays, tables and model workflowsTransfer learning vs fine-tuningData leakage in machine learningBuild the foundations for this subject
Follow SchoolWhool's learning path through Python, Data Tools, Maths, Statistics and Machine Learning. Start with the skills you need, read open lessons without an account, then sign in free for quizzes and saved progress. Free sample modules also open their practice and study material.
