Data exploitation: multi-task learning of object detection and semantic segmentation on partially annotated data
Hoàng-Ân Lê (IRISA, University of South Brittany),* Minh-Tan Pham (IRISA-UBS)
The 34th British Machine Vision Conference
Abstract
Multi-task partially annotated data, where each data point has annotations for only single task, is potentially helpful for data scarcity. In this paper, we study the joint learning of object detection and semantic segmentation, the most two popular vision problems, from multi-task data with partial annotations. Extensive experiments are performed to evaluate each task performance and explore their complementarity when a multi-task network cannot optimize both tasks simultaneously. A simple knowledge distillation method is proposed as an alternative for joint-task optimization. All code and data splits will be publicly released upon acceptance to facilitate further research.Video
Citation
@inproceedings{Lê_2023_BMVC,
author = {Hoàng-Ân Lê and Minh-Tan Pham},
title = {Data exploitation: multi-task learning of object detection and semantic segmentation on partially annotated data},
booktitle = {34th British Machine Vision Conference 2023, {BMVC} 2023, Aberdeen, UK, November 20-24, 2023},
publisher = {BMVA},
year = {2023},
url = {https://papers.bmvc2023.org/0870.pdf}
}
Copyright © 2023
The British Machine Vision Conference is organised by