PanoMitoAtlas

The dataset includes 2D images, time-lapse images, and their corresponding annotations in COCO format, as well as the training/test splits used in PanoMito, enabling researchers to reproduce our experiments and reuse the dataset for future studies. The PanoMitoAtlas.zip archive contains the following components: all_2d_dataset: This folder contains 6,481 mitochondrial images and 131,940 corresponding annotations. The annotation file is provided as _all_data_annotations.json. timeline: This folder contains 71 time-lapse mitochondrial images and 858 corresponding annotations. The annotation file is provided as Timeline_annotations_all_rle_timeline_id.json. train_test_subset: This folder contains the predefined training and test subsets used in our manuscript. The train folder contains 1,422 mitochondrial images, 51,781 corresponding annotations in _train_subdataset_gt.json, and the RLE-encoded annotation file used for training in NumPy format, _train_subdataset_gt_nocate_rle.npy. The test folder contains 611 mitochondrial images, 21,928 corresponding annotations in _test_subdataset_gt.json, and the RLE-encoded annotation file used for evaluation/training-format loading in NumPy format, _test_subdataset_gt_nocate_rle.npy.