Semantic Segmentation

Semantic segmentation is a computer vision task that involves classifying each pixel in an image into predefined categories, effectively partitioning the image into regions corresponding to different classes. This task can be approached using convolutional neural networks (CNNs) that output a dense prediction map, where each pixel is assigned a class label. The architecture typically employs an encoder-decoder structure, such as U-Net or Fully Convolutional Networks (FCNs), to capture both global context and local details. The loss function commonly used in semantic segmentation is the pixel-wise cross-entropy loss, which measures the discrepancy between predicted and true labels. Evaluation metrics include pixel accuracy, mean Intersection over Union (IoU), and class-wise IoU, which provide insights into the model's performance across different categories.