Fluorescence Microscopy Images for Cellular Detection with Bounding Box Annotations

Six distinct collections of fluorescence microscopy images accompany a paper on object detection networks and augmented reality for cellular analysis. The data includes 80 to 180 annotated images per subset, featuring erythroblast, neuroblastoma, fibroblast, eukaryote, C127, and Hela cells stained with various markers. Dominic Waithe from the University of Oxford created this dataset, which contains bounding box annotations compatible with Faster-RCNN and YOLOv2.

Use Cases

• Training object detection networks like Faster-RCNN based on bounding box annotations for cellular structures.
• Benchmarking model performance on different cell types and stains mentioned, such as DAPI and phalloidin.
• Developing augmented reality tools for microscopy acquisition and analysis as referenced in the paper's title.

Strengths

• Contains six distinct biological subsets with specific staining protocols detailed in the description.
• Annotations were created by a skilled user and converted into formats compatible with major object detection frameworks.
• Provides exact image counts for training and testing per subset, such as 180 training and 180 testing images for neuroblastoma cells.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.
• The total dataset size and file formats are unspecified, complicating download and storage planning.

Provenance

Source

University of Oxford, accompanying a specific research paper.

Collection Method

Microscopy images were acquired from lab staining and the Cell Image Library, then annotated with bounding boxes.