Langenkämper, Daniel ; van Kevelaer, Robin; Purser, Autun; Nattkemper, Tim Wilhelm: Gear-Induced Concept Drift in Marine Images and Its Effect on Deep Learning Classification. In: Frontiers in Marine Science. Jg.7. 2020
Inhalt
- 2.1. Image Data Sets
- 2.2. Concept Drift Visualization With t-SNE Projections
- 2.3. ResNet Deep Learning Architecture and Training
- 2.4. Study Design
- 2.4.1. Experiment A: Intra-set Study (No Concept Drift)
- 2.4.2. Experiment B: 1 vs. 1 Inter-set Study
- 2.4.3. Experiment C: Leave-One-Set-Out Inter-set Study
- 2.4.4. Experiment D: Ensemble Classification Heuristic
- 2.4.5. Experiment E: Leave-One-Set-Out With Adaption
- 2.4.6. Balanced Data Set Experiments
- 3. Results
- 4. Discussion
- 5. Conclusion
- Data Availability Statement
- Author Contributions
- Funding
- Acknowledgments
- Supplementary Material
- References