Dataset Viewer
Auto-converted to Parquet Duplicate
sequence
stringlengths
25
68
preset
stringclasses
8 values
tier
stringclasses
3 values
condition
stringclasses
2 values
seed
int64
0
20
size_GiB
float64
0.16
11.2
overrides
stringlengths
75
129
a549_filopodia/a549_filopodia_easy_s0
a549_filopodia
easy
treated
0
1.6
scene.n_cells=1;acquisition.photons_per_unit=110;render.voxel_um=[0.125,0.125,1]
a549_filopodia/a549_filopodia_medium_s0
a549_filopodia
medium
treated
0
1.59
scene.n_cells=1;acquisition.photons_per_unit=55;render.voxel_um=[0.125,0.125,1]
a549_filopodia/a549_filopodia_hard_s0
a549_filopodia
hard
treated
0
0.79
scene.n_cells=1;acquisition.photons_per_unit=22;render.voxel_um=[0.125,0.125,2]
apoptosis_hep2/apoptosis_hep2_easy_s0
apoptosis_hep2
easy
treated
0
3.81
scene.n_cells=12;acquisition.photons_per_unit=120;render.voxel_um=[0.5,0.5,1]
apoptosis_hep2/apoptosis_hep2_medium_s0
apoptosis_hep2
medium
treated
0
3.83
scene.n_cells=20;acquisition.photons_per_unit=60;render.voxel_um=[0.5,0.5,1]
apoptosis_hep2/apoptosis_hep2_hard_s0
apoptosis_hep2
hard
treated
0
1.93
scene.n_cells=30;acquisition.photons_per_unit=24;render.voxel_um=[0.5,0.5,2]
apoptosis_hep2/apoptosis_hep2_medium_control_s3
apoptosis_hep2
medium
control
3
3.9
scene.n_cells=20;acquisition.photons_per_unit=60;render.voxel_um=[0.5,0.5,1]
emt_colony2d/emt_colony2d_easy_induction_strength2.0_s3
emt_colony2d
easy
treated
3
0.46
scene.n_cells=72;acquisition.photons_per_unit=120;render.voxel_um=[1.0,1.0,2];emt.induction_strength=2.0
emt_colony2d/emt_colony2d_medium_induction_strength2.0_s3
emt_colony2d
medium
treated
3
0.46
scene.n_cells=120;acquisition.photons_per_unit=60;render.voxel_um=[1.0,1.0,2];emt.induction_strength=2.0
emt_colony2d/emt_colony2d_hard_induction_strength2.0_s3
emt_colony2d
hard
treated
3
0.21
scene.n_cells=180;acquisition.photons_per_unit=24;render.voxel_um=[1.0,1.0,4];emt.induction_strength=2.0
emt_colony2d/emt_colony2d_medium_control_s4
emt_colony2d
medium
control
4
0.47
scene.n_cells=120;acquisition.photons_per_unit=60;render.voxel_um=[1.0,1.0,2]
emt_lumenoid/emt_lumenoid_easy_induction_strength2.0_s15
emt_lumenoid
easy
treated
15
4.68
scene.n_cells=54;acquisition.photons_per_unit=120;render.voxel_um=[0.5,0.5,2];emt.induction_strength=2.0
emt_lumenoid/emt_lumenoid_medium_induction_strength2.0_s15
emt_lumenoid
medium
treated
15
4.76
scene.n_cells=90;acquisition.photons_per_unit=60;render.voxel_um=[0.5,0.5,2];emt.induction_strength=2.0
emt_lumenoid/emt_lumenoid_hard_induction_strength2.0_s15
emt_lumenoid
hard
treated
15
2.39
scene.n_cells=135;acquisition.photons_per_unit=24;render.voxel_um=[0.5,0.5,4];emt.induction_strength=2.0
emt_lumenoid/emt_lumenoid_medium_control_s20
emt_lumenoid
medium
control
20
4.79
scene.n_cells=90;acquisition.photons_per_unit=60;render.voxel_um=[0.5,0.5,2]
hl60_mitosis/hl60_mitosis_easy_s0
hl60_mitosis
easy
treated
0
10.78
scene.n_cells=24;acquisition.photons_per_unit=360;render.voxel_um=[0.125,0.125,0.2]
hl60_mitosis/hl60_mitosis_medium_s0
hl60_mitosis
medium
treated
0
11.16
scene.n_cells=40;acquisition.photons_per_unit=180;render.voxel_um=[0.125,0.125,0.2]
hl60_mitosis/hl60_mitosis_hard_s0
hl60_mitosis
hard
treated
0
5.6
scene.n_cells=60;acquisition.photons_per_unit=72;render.voxel_um=[0.125,0.125,0.4]
mda231_invasion/mda231_invasion_easy_s0
mda231_invasion
easy
treated
0
1.34
scene.n_cells=7;acquisition.photons_per_unit=50;render.voxel_um=[0.5,0.5,2]
mda231_invasion/mda231_invasion_medium_s0
mda231_invasion
medium
treated
0
1.36
scene.n_cells=12;acquisition.photons_per_unit=25;render.voxel_um=[0.5,0.5,2]
mda231_invasion/mda231_invasion_hard_s0
mda231_invasion
hard
treated
0
0.68
scene.n_cells=18;acquisition.photons_per_unit=10;render.voxel_um=[0.5,0.5,4]
pgc_bleb/pgc_bleb_easy_s0
pgc_bleb
easy
treated
0
0.34
scene.n_cells=2;acquisition.photons_per_unit=50;render.voxel_um=[0.5,0.5,2]
pgc_bleb/pgc_bleb_medium_s0
pgc_bleb
medium
treated
0
0.35
scene.n_cells=3;acquisition.photons_per_unit=25;render.voxel_um=[0.5,0.5,2]
pgc_bleb/pgc_bleb_hard_s0
pgc_bleb
hard
treated
0
0.16
scene.n_cells=4;acquisition.photons_per_unit=10;render.voxel_um=[0.5,0.5,4]
sars2_syncytia/sars2_syncytia_easy_moi0.3_tmprss2_multiplier4.0_s7
sars2_syncytia
easy
treated
7
3.89
scene.n_cells=120;acquisition.photons_per_unit=60;render.voxel_um=[0.65,0.65,2];virus.moi=0.3;virus.fusion.tmprss2_multiplier=4.0
sars2_syncytia/sars2_syncytia_medium_moi0.3_tmprss2_multiplier1.0_s6
sars2_syncytia
medium
treated
6
3.95
scene.n_cells=200;acquisition.photons_per_unit=30;render.voxel_um=[0.65,0.65,2];virus.moi=0.3;virus.fusion.tmprss2_multiplier=1.0
sars2_syncytia/sars2_syncytia_hard_moi0.3_tmprss2_multiplier4.0_s7
sars2_syncytia
hard
treated
7
1.68
scene.n_cells=300;acquisition.photons_per_unit=12;render.voxel_um=[0.65,0.65,4];virus.moi=0.3;virus.fusion.tmprss2_multiplier=4.0
sars2_syncytia/sars2_syncytia_medium_control_s8
sars2_syncytia
medium
control
8
4.28
scene.n_cells=200;acquisition.photons_per_unit=30;render.voxel_um=[0.65,0.65,2]

cellgen — subset

Simulated 3D+t fluorescence microscopy of cells, with dense ground truth on every frame. This repository holds a subset of 28 sequences (8 cell systems × 3 difficulty tiers, plus one negative control per perturbed system) out of the full 241-sequence benchmark.

Status: private preview. License to be decided before public release.

Download

hf download Tommy1118/cellgen-subset --repo-type dataset --local-dir cellgen_subset
cd cellgen_subset
zstd -d -c cellgen_subset.tar.zst | tar -xf -        # ~87 GB extracted

Contents

system phenomenon scene frames × Δt sequences
hl60_mitosis HL-60: mitosis, crowding suspension, 40 cells 144 × 5 min easy, medium, hard
a549_filopodia A549: filopodia (number, length, width, branching, lifetime) single cell 120 × 2 min easy, medium, hard
mda231_invasion MDA-MB-231: strong deformation, pseudopodia, invasion 3D matrix, 12 cells 120 × 5 min easy, medium, hard
pgc_bleb zebrafish PGC: blebs at conserved volume 3D matrix, 3 cells 180 × 5 s easy, medium, hard
apoptosis_hep2 Hep-2 + etoposide: apoptosis (AVD, blebbing, condensation, fragmentation), 10% necrosis monolayer, 20 cells 240 × 2 min easy, medium, hard + control
emt_lumenoid hiPSC: EMT, basement-membrane crossing, lumen collapse (induction strength 2.0) lumenoid, 90 cells 120 × 30 min easy, medium, hard + control
emt_colony2d hiPSC: EMT, scattering from the colony (induction strength 2.0) 2D colony, 120 cells 120 × 30 min easy, medium, hard + control
sars2_syncytia U2OS-ACE2: infection, fusion into syncytia, syncytium death (MOI 0.3) monolayer, 200 cells 216 × 10 min easy, medium, hard + control

manifest.tsv lists every sequence with its tier, condition, seed, size and the configuration overrides it was generated with.

Tiers. Relative to medium, easy has 0.6× the cell density and 2× the photon budget; hard has 1.5× the density, 0.4× the photons and twice the z step.

Selection. Per system and tier, the treated sequence with the strongest perturbation and the lowest seed; for perturbed systems, also the medium-tier control with the lowest seed. Every sequence passes the label consistency check (no gaps inside a track, valid lineage links, SEG and TRA identical). Virus cytopathic-effect presets (vaccinia, HSV-1, rhinovirus) are not included.

Layout of one sequence

<system>/<sequence>/
  raw.ome.zarr/        noisy 16-bit stack (t, c, z, y, x), 2x xy pyramid, labels/{nuclei,cells}
  clean.ome.zarr/      noise-free expected detector signal (same grid)
  nuc/                 Cell Tracking Challenge N-series: 01/tNNN.tif, 01_GT/{SEG,TRA}, man_track.txt
  cell/                Cell Tracking Challenge C-series (whole cells), same layout (absent for hl60_mitosis)
  STATE/cells.csv      per cell and frame: phase, death stage, EMT state, infection, syncytium,
  STATE/nuclei.csv       number of nuclei, true volume, surface area, sphericity, inside basement membrane
  STATE/events.csv     every division, death, transition, infection and fusion
  difficulty.json      CTC dataset descriptors (SNR, contrast ratio, heterogeneity, size, spacing,
                       contact, density, division rate, signal change)
  meta.json            voxel size, frame interval, PSF, track and event counts
  config.yaml          full resolved simulation configuration

The pre-PSF density volume (phantom.ome.zarr) of the full release is not included here.

Conventions. Labels follow the Cell Tracking Challenge format. A cell absorbed into a syncytium ends its cell track (event fusion in events.csv), while its nucleus keeps its own track inside the host; CTC labels cannot express merges, so events.csv is the record. A nucleus that leaves the field of view and returns starts a new track with parent 0 (01_GT/TRA/reentry_map.txt).

Simulation

Each sequence is produced in three stages: (i) an agent model updates position, shape, cell-cycle phase, mitosis and the state machine of each phenomenon; (ii) a rasterizer turns every cell and nucleus into an implicit field on a fine isotropic grid, assigning each voxel to the strongest field so that cells never overlap, with volume conserved frame to frame; (iii) an imaging model convolves with a depth-invariant PSF (vectorial confocal, widefield for HL-60), samples z planes, integrates over pixels, and applies photobleaching, background photons, Poisson shot noise, Gaussian read noise, a nonuniform camera offset and 16-bit quantization. HL-60 and A549 are calibrated against Fluo-N3DH-SIM+ 01 and Fluo-C3DH-A549 01 by median matching; the other systems use literature parameter ranges.

Downloads last month
39