Viet-Anh on Software Logo

What is: Models Genesis?

SourceModels Genesis
Year2000
Data SourceCC BY-SA - https://paperswithcode.com

Models Genesis, or Generic Autodidactic Models, is a self-supervised approach for learning 3D image representations. The objective of Models Genesis is to learn a common image representation that is transferable and generalizable across diseases, organs, and modalities. It consists of an encoder-decoder architecture with skip connections in between, and is trained to learn a common image representation by restoring the original sub-volume x_ix\_{i} (as ground truth) from the transformed one xˉ_i\bar{x}\_{i} (as input), in which the reconstruction loss (MSE) is computed between the model prediction x_0x'\_{0} and ground truth x_ix\_{i}. Once trained, the encoder alone can be fine-tuned for target classification tasks; while the encoder and decoder together can be fine-tuned for target segmentation tasks.