IAUNet: Instance-Aware U-Net
Keywords: Medical and Biological Vision, Cell Microscopy, Instance Segmentation, Deep Learning
Abstract: Instance segmentation is critical in biomedical imaging for accurately distinguishing individual objects, such as cells, which often overlap and vary in size. Recent query-based methods—where object-specific queries guide segmentation—have shown strong performance in this task. While U-Net has been a go-to architecture in medical image segmentation, it was neither specifically designed for instance segmentation nor explored in the context of query-based approaches. In this work, we present IAUNet, a novel architecture that brings instance awareness to U-Net with query-based mechanisms to achieve superior pixel-to-instance clustering. The key design includes lightweight Instance Activation (IA) layers, which generate guided object queries by highlighting semantically important regions. Additionally, we propose a Parallel Dual-Path Transformer decoder that refines object-specific features across multiple scales, allowing us to assign multiple queries from different scale levels to a specific object. Finally, we introduce the 2025 Revvity Full Cell Segmentation Dataset, comprising hundreds of manually labeled cells from brightfield images. This dataset is unique in capturing the complex morphology of overlapping cell cytoplasm with an unprecedented level of detail, making it a valuable resource and benchmark for advancing instance segmentation in biomedical imaging. Experiments on multiple public datasets and our own show that IAUNet outperforms most state-of-the-art fully convolutional, transformer-based, and query-based models, setting a strong baseline for medical image instance segmentation tasks.
Primary Area: applications to computer vision, audio, language, and other modalities
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 12188
Loading