Reading Diary Week 2

5 Retinal Vessel Segmentation terms

Data-augmentation: It is a technique applied when there are small data during the experiment to increase the size and diversity of samples.; Example: Although data augmentation is performed for the original datasets, serious overfitting is still observed.
Downsampling: It is applying convolutional layer followed by a max-pooling layer which eventually reduces the size of the data and increases the number of channels.; Example: Basically, U-Net consists of a typical downsampling encoder and upsampling decoder structure.
Lightweight-Network: It is a part of a convolutional neural network designed by compressing and reducing the weight from the original network.; Example: In this work, we propose a lightweight-network named Spatial Attention U-Net.
Overfitting: It means when a machine learning is good at predicting the data it has seen before but bad at new data.; Example: Although data augmentation is performed for the original datasets, serious overfitting is still observed.
Upsampling: It is applying transposed convolutional layer which eventually increases the size of the data and decreases the number of channels.; Example: Basically, U-Net consists of a typical downsampling encoder and upsampling decoder structure.