Leveraging knowledge distillation for partial multi-task learning from multiple remote sensing datasets
This work addresses a specific challenge in remote sensing by enabling more efficient multi-task learning from disparate datasets, though it is incremental as it builds on existing knowledge distillation techniques.
The paper tackles the problem of partial multi-task learning in remote sensing, where datasets have annotations for only one task each, by using knowledge distillation to improve joint representation learning without needing all-task ground truths, resulting in demonstrated effectiveness on semantic tasks like object detection and segmentation in aerial images.
Partial multi-task learning where training examples are annotated for one of the target tasks is a promising idea in remote sensing as it allows combining datasets annotated for different tasks and predicting more tasks with fewer network parameters. The naïve approach to partial multi-task learning is sub-optimal due to the lack of all-task annotations for learning joint representations. This paper proposes using knowledge distillation to replace the need of ground truths for the alternate task and enhance the performance of such approach. Experiments conducted on the public ISPRS 2D Semantic Labeling Contest dataset show the effectiveness of the proposed idea on partial multi-task learning for semantic tasks including object detection and semantic segmentation in aerial images.