Completed
Semi-Supervised Learning: Self-Tr
Class Central Classrooms beta
YouTube videos curated by Class Central.
Classroom Contents
Data-Efficient AI for Accelerating MRI Acquisition
Automatically move to the next video in the Classroom when playback concludes
- 1 Intro
- 2 Magnetic Resonance Imaging
- 3 Data Pipeline
- 4 Raw Data in MRI
- 5 k-space Sampling
- 6 Constraints
- 7 Can We Speed Things Up??
- 8 MRI Super-Resolution Convert low resolution (LR) to high resolution (HR)
- 9 Domain Knowledge • Embedding physical principles into the model
- 10 Reducing Data Requirements Domain knowledge reduce extent of ill-posed inverse problems
- 11 What if we have no training data
- 12 Deep Image Prior
- 13 Convolutional Decoder ConvDe
- 14 Semi-Supervised Learning: Self-Tr
- 15 Self-Training for MRI Recon
- 16 Why is ConvDecoder Better? Zero Unrolled ConvDecoder
- 17 ConvDecoder Noisy Student
- 18 Self-Training Takeaway Keep forward model identical; modify pseudo-labels
- 19 Invariance to Forward Model Change Supervised Unsupervised
- 20 Generalized Consistency Framew
- 21 Artifact Invariant Reconstruction
- 22 Metrics for Evaluation Discordance between quantitative and qualitative metrics
- 23 New Multi-Task Dataset
- 24 Surrogates without Dense Labels Use self-supervised tasks to build image representations
- 25 Self-Supervised Image Quality
- 26 Overall Takeaway • Adding domain knowledges reduces data requirements
- 27 Open Source Data/Code
- 28 Conclusions
