Main Idea of the Document: The document discusses the importance of balancing network width, depth, and resolution in Convolutional Neural Networks (ConvNets) to achieve better accuracy and efficiency. It introduces a compound scaling method that allows scaling up a baseline ConvNet effectively to meet different resource constraints while maintaining model efficiency. This method is demonstrated through the example of scaling a mobile-size EfficientNet model to surpass state-of-the-art accuracy with fewer parameters and FLOPS.
Sources: p. 1