Computer Vision
Kaggle Learn
A short, code-first course on building convolutional neural networks with TensorFlow and Keras. It takes a convolutional classifier apart one mechanism per lesson: the convolution operation and the rectified linear unit (ReLU) activation that follows it, maximum pooling, and the stride and padding settings that control how a filter sweeps across an image. Later lessons have you design a network of your own and use data augmentation — creating extra training data — to get more out of a small dataset. Six lessons, each a tutorial followed by a hands-on exercise, take about four hours. Optional bonus material covers transfer learning, a cassava leaf-disease classification problem, and running on Kaggle's TPUs for the Petals to the Metal competition. It continues from Kaggle's Intro to Deep Learning course.
What you'll learn
- How a convolutional classifier is assembled, from feature extraction to the classifier head
- The convolution operation and the rectified linear unit (ReLU) activation that follows it
- Maximum pooling, and what it throws away to make features stand out
- Stride and padding — how the filter window moves across an image
- Designing a convolutional network of your own
- Data augmentation: creating extra training data to boost performance
Frequently asked questions about Computer Vision
Who is Computer Vision for?
Developers with some Python and neural-network experience who want to understand how convolutional networks classify images, and build one.
Is Computer Vision free?
Yes — Computer Vision is completely free to take.
What are the prerequisites for Computer Vision?
Python, plus the neural-network and Keras basics covered by Kaggle's Intro to Deep Learning course.
Do you need to code for Computer Vision?
Yes — Computer Vision involves hands-on coding.
Does Computer Vision offer a certificate?
Yes. A Kaggle Learn certificate of completion, earned by finishing the six core lessons and their exercises. The bonus lessons are optional and not needed for it.
Why we suggest this course
For a developer who has trained a basic neural network and wants to understand how a vision model actually sees, this pulls the classifier apart and rebuilds it, with code you run yourself and an exercise after every lesson. The bonus lessons are a genuine extra rather than filler — transfer learning and a real Kaggle competition entry. One thing to know: it starts where Kaggle's Intro to Deep Learning course finishes, so neural-network and Keras fundamentals are assumed rather than taught; start there if they are new.