TY - GEN
T1 - JLInst
T2 - 6th Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2023
AU - Zhao, Xiaodong
AU - Chen, Junliang
AU - Huang, Zepeng
AU - Shen, Linlin
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd 2024.
PY - 2024
Y1 - 2024
N2 - Lots of methods have been proposed to improve instance segmentation performance. However, the mask produced by state-of-the-art segmentation networks is still coarse and does not completely align with the whole object instance. Moreover, we find that better object boundary information can help instance segmentation network produce more distinct and clear object masks. Therefore, we present a simple yet effective instance segmentation framework, termed JLInst (Boundary-Mask Joint Learning for Instance Segmentation). Our methods can jointly exploit object boundary and mask semantic information in the instance segmentation network, and generate more precise mask prediction. Besides, we propose the Adaptive Gaussian Weighted Binary Cross-Entropy Loss (GW loss), to focus more on uncertain examples in pixel-level classification. Experiments show that JLInst achieves improved performance (+3.0% AP) than Mask R-CNN on COCO test-dev2017 dataset, and outperforms most recent methods in the fair comparison.
AB - Lots of methods have been proposed to improve instance segmentation performance. However, the mask produced by state-of-the-art segmentation networks is still coarse and does not completely align with the whole object instance. Moreover, we find that better object boundary information can help instance segmentation network produce more distinct and clear object masks. Therefore, we present a simple yet effective instance segmentation framework, termed JLInst (Boundary-Mask Joint Learning for Instance Segmentation). Our methods can jointly exploit object boundary and mask semantic information in the instance segmentation network, and generate more precise mask prediction. Besides, we propose the Adaptive Gaussian Weighted Binary Cross-Entropy Loss (GW loss), to focus more on uncertain examples in pixel-level classification. Experiments show that JLInst achieves improved performance (+3.0% AP) than Mask R-CNN on COCO test-dev2017 dataset, and outperforms most recent methods in the fair comparison.
KW - Boundary Information Enhancement
KW - Instance Segmentation
UR - http://www.scopus.com/inward/record.url?scp=85181984736&partnerID=8YFLogxK
U2 - 10.1007/978-981-99-8555-5_35
DO - 10.1007/978-981-99-8555-5_35
M3 - Conference contribution
AN - SCOPUS:85181984736
SN - 9789819985548
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 445
EP - 457
BT - Pattern Recognition and Computer Vision - 6th Chinese Conference, PRCV 2023, Proceedings
A2 - Liu, Qingshan
A2 - Wang, Hanzi
A2 - Ji, Rongrong
A2 - Ma, Zhanyu
A2 - Zheng, Weishi
A2 - Zha, Hongbin
A2 - Chen, Xilin
A2 - Wang, Liang
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 13 October 2023 through 15 October 2023
ER -