Pre-processing training data improves accuracy and generalisability of convolutional neural network based landscape semantic segmentation.
Andrew ClarkStuart R. PhinnPeter F. ScarthPublished in: CoRR (2023)
Keyphrases
- semantic segmentation
- training data
- street scenes
- conditional random fields
- scene classification
- superpixels
- data sets
- label transfer
- weakly supervised
- feature extraction
- training set
- learning algorithm
- supervised learning
- decision trees
- training examples
- training samples
- co occurrence
- feature space
- object recognition