Helmholtz Gemeinschaft

Search
Browse
Statistics
Feeds

FISBe: a real-world benchmark dataset for instance segmentation of long-range thin filamentous structures

Item Type:Conference or Workshop Item
Title:FISBe: a real-world benchmark dataset for instance segmentation of long-range thin filamentous structures
Creators Name:Mais, L., Hirsch, P., Managan, Claire, Kandarpa, Ramya, Rumberger, J.L., Reinke, Annika, Maier-Hein, Lena, Ihrke, Gudrun and Kainmueller, D.
Abstract:Instance segmentation of neurons in volumetric light microscopy images of nervous systems enables ground-breaking research in neuroscience by facilitating joint functional and morphological analyses of neural circuits at cel-lular resolution. Yet said multi-neuron light microscopy data exhibits extremely challenging properties for the task of instance segmentation: Individual neurons have long-ranging, thin filamentous and widely branching morpholo-gies, multiple neurons are tightly inter-weaved, and par-tial volume effects, uneven illumination and noise inherent to light microscopy severely impede local disentan-gling as well as long-range tracing of individual neurons. These properties reflect a current key challenge in machine learning research, namely to effectively capture long-range dependencies in the data. While respective method-ological research is buzzing, to date methods are typically benchmarked on synthetic datasets. To address this gap, we release the FlyLight Instance Segmentation Benchmark (FISBe) dataset, the first publicly available multi-neuron light microscopy dataset with pixel-wise annotations. In addition, we define a set of instance segmentation metrics for benchmarking that we designed to be meaningful with regard to downstream analyses. Lastly, we provide three baselines to kick off a competition that we envision to both advance the field of machine learning regarding methodology for capturing long-range data dependencies, and facilitate scientific discovery in basic neuroscience.
Keywords:Instance Segmentation, Dataset, Evaluation, Long-Range Data Dependencies, Neuroscience, 3D Light Microscopy
Source:2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
ISSN:1063-6919
ISBN:979-8-3503-5300-6
Publisher:IEEE
Page Range:22249-22259
Number of Pages:11
Date:16 September 2024
Official Publication:https://doi.org/10.1109/CVPR52733.2024.02100

Repository Staff Only: item control page

Open Access
MDC Library