Lecture 10 - Recognition Flashcards
What is image classification?
Image classification is the task of assigning a label from a predefined set of categories to an entire image.
Describe the architecture of LeNet.
LeNet is one of the first convolutional networks designed for handwritten digit recognition, featuring layers of convolutional filters followed by subsampling (pooling) and fully connected layers.
What is AlexNet and its significance?
AlexNet is a convolutional network that won the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) in 2012, significantly outperforming previous models with its deep architecture and use of ReLU activation and dropout.
Explain dropout in neural networks.
Dropout is a regularization technique where a random subset of neurons is ignored during training, preventing overfitting by ensuring the model does not rely too heavily on any single neuron.
What is data augmentation and its purpose?
Data augmentation artificially enlarges the training dataset by applying label-preserving transformations such as cropping, flipping, and color jittering to improve model generalization.
Describe the purpose of batch normalization.
Batch normalization normalizes the inputs of each layer to have a mean of zero and variance of one, accelerating training and improving model performance by reducing internal covariate shift.
What are the key features of VGGNet?
VGGNet uses small 3x3 convolutional filters stacked together, leading to deep networks with more non-linearities, which improve performance while maintaining computational efficiency.
Explain the Inception module in GoogLeNet.
The Inception module in GoogLeNet combines multiple convolutional filters of different sizes and pooling operations within the same layer, allowing the network to capture features at various scales.
What is a Residual Network (ResNet)?
ResNet introduces residual blocks with shortcut connections that bypass one or more layers, allowing very deep networks to be trained by mitigating the vanishing gradient problem.
Describe the concept of feature learning in CNNs.
Feature learning in CNNs refers to the automatic extraction and learning of hierarchical features from raw input images, replacing the need for manual feature engineering.
What is the role of the softmax function in CNNs?
The softmax function is used in the output layer of CNNs for multi-class classification, converting raw output scores into probabilities by normalizing them.
Explain the concept of transfer learning in CNNs.
Transfer learning involves using a pre-trained network on a large dataset and fine-tuning it on a smaller, task-specific dataset, leveraging the learned features to improve performance and reduce training time.
Write the formula for the softmax function.
Provide the formula for the ReLU activation function.
What is the formula for updating weights using gradient descent in neural networks?