What is the difference between stacked autoencoders and CNN for pixel-wise classification based feature extraction ? Is that the quality of the features, the processing time ?
Stacked autoencoders can be trained on unlabeled data, whereas pixel-wise classifiers like FCNs can not.
Stacked auto-encoders are unsupervised models, while CNNs are supervised models. If your data is labeled, you should use CNN for better results.
Stacked auto encoder is a method for unsupervised learning and CNN of course supervised. But for good classification results CNN is better way.
If you don't have labeled data then SAE is best option and if you have labeled data then depends on your data nature which technique would be suitable for pixel wise classification.
- Badges
- Science topic
- Similar topics
- Analytical Chemistry
- Extraction