Dataset Viewer
Auto-converted to Parquet Duplicate
text
stringlengths
0
122
frames: 900
frame_rate_hz: 15
height: 512
width: 512
bit_depth: 8
micrometres_per_pixel: 1.5
duration_s: 60
Ground truth for the training movie. Spatial arrays are in pixel coordinates of the zero-motion frame, (y, x) order.
cells.npz centers (n, 2) float; radius_px (n,); in_fov (n,) bool, True for the cells whose centre lies inside
the field of view (the others sit in the margin and only poke in); pnr (n,) peak-to-noise ratio of
the ROI-mean trace, camera noise only
footprints_sparse.npz scipy CSR matrix (n, height * width) of the nominal spatial footprints, peak about 1;
load with scipy.sparse.load_npz and reshape each row to (height, width)
traces.npz spikes (n, T) counts per frame; dff (n, T) the somatic dF/F; fluorescence (n, T) the somatic
fluorescence in arbitrary units; the rendered cell is footprint x fluorescence before optics,
motion and the camera

Calcium source extraction

Synthetic one-photon (miniscope-style) calcium imaging of mouse cortex, rendered with a physically motivated simulator that models somatic morphology at several depths, calcium indicator kinetics, correlated neuropil, vasculature and hemodynamics, non-stationary rigid motion, wide-field optics and an sCMOS camera, with complete ground truth. The simulator is not part of this release.

data/heldout/video.tif        60 s at 15 Hz, 512 x 512, 8-bit, 900 frames: the movie to extract from
data/heldout/acquisition.txt  frame rate, pixel size, bit depth
data/train/video.tif          a second movie from the same configuration and a different seed
data/train/ground_truth/      its cells, footprints and traces, described in README.txt there

The held-out movie's ground truth is not published.

Licence

CC BY 4.0. Attribute as "Calcium source extraction, Emulated".

Downloads last month
8