EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks - Google Slides

EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks   Slideshow     Share FileEditViewHelp   View only 1 EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks Mingxin Tan, Quoc V. Le Presented by: Jesse Sun 2 Outline Part II VGG, ResNet, DenseNet, SqueezeNet, MobileNet Part IV Experiments and Results Part I Quick Intro to Convolutional Neural Networks Part V Conclusion and Discussion Part III EfficientNets 3 Part I: Quick Intro to Convolutional Neural Networks 4 Convolutions ● Have existed for a long time. ● Mathematical operation on two functions f and g which products a third function that describes how the shape of one is modified by the other. ● Used in many areas like digital signal processing, and more recently with machine learning. Part I: Quick Intro to Convolutional Neural Networks 5 As with all good ideas in machine learning... They “always” stem from Geoffrey Hinton. 6 AlexNet (Alex Krizhevsky, Ilya Sutskever, Geoffrey Hinton 2012) ●

saved by

• 3017