Academic Gallery
Delve has been used in several papers:
ResNet18 trained on Cifar10 for 30 epochs using the adam optimizer and a batch size of 64. Image from “Should You Go Deeper? Optimizing Convolutional Neural Networks without training”.
ResNet34 trained on Cifar10 for 30 epochs using the adam optimizer. Image from “Should You Go Deeper? Optimizing Convolutional Neural Networks without training”.
DenseNet18 trained on Food101 for 90 epochs using the stochastic gradient decent optimizer and a batch size of 128. Image from “Feature Space Saturation During Training”.
VGG16 trained on 3 different resolutions for 30 epochs using the Adam-optimizer and a batch size of 32. You can see the shift in the inference process by observing the shift in high saturation values. Image from “(Input) Size Matters for Convolutional Neural Network Classifiers”.