← All modelsarchitecture
Convolutional Neural Network (CNN)
intermediatesupervised · self_supervised
parametricclassificationrepresentationimagevideo
A CNN stacks convolutional layers to turn pixels into features and, ultimately, a prediction.
Mechanisms
backpropagationnonlinear activationdense layers
Properties
nonlinearrepresentation learningpretrained ecosystem
Constraints
high computerequires large datasensitive to tuning
Practical profile
- Explainability
- low
- Training cost
- high
- Inference cost
- medium
- Data appetite
- high
