AMAR POTDAR Thoughts, Experiences, Projects

CNN Series: ResNet

Paper Deep Residual Learning for Image Recognition Authors: Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun ResNet ResNet short for Residual Network it tackled a major challenge faced by deep convolutional neural networks (CNNs) callws vanishing gradients. As CNNs grew deeper (more layers), training became increasi... Read more

CNN Series: VGGNet

Paper Very Deep Convolutional Networks for Large-Scale Image Recognition Authors: Karen Simonyan, Andrew Zisserman VGGNet VGGNet are series of Convolutional Neural Network models proposed by the Visual Geometry Group of Oxford University, including the well known VGG11,VGG13,VGG16 and VGG19. Simplicity of Architecture: VGGNet’s arch... Read more

CNN Series: AlexNet

Paper ImageNet Classification with Deep Convolutional Neural Networks Authors: Alex Krizhevsky, Ilya Sutskever, Geoffrey E. Hinton AlexNet AlexNet is a convolutional neural network that won the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) in 2012. It was designed to recognize objects and faces from over a million differe... Read more

LearningByDoing Series: Building Blocks of Neural Network - Activation Functions

Activation Functions: Purpose: Activation functions determine the output of a neural network node given an input or set of inputs, introducing non-linear properties to the network enabling them to learn complex patterns. Activation functions help decide whether a neuron should be activated or not, influencing the network’... Read more

LearningByDoing Series: Understanding effects of activation functions

Ablation study: Activation functions Goal here is to observ or record changes in accuracy, loss, or other performance indicators. Activation Functions under consideration Sigmoid Tanh ReLU LeakyReLU ELU SELU Softplus Mish GELU Swish We are using ResNet-18 model wit... Read more