# ZAPBench: Zebrafish Whole-Brain Activity Prediction Benchmark

## Facts

- Provider: Google Research and HHMI Janelia (Ahrens lab, CellMap Project Team)
- License: CC-BY-4.0 (https://storage.googleapis.com/zapbench-release/volumes/README.md)
- Format: PNG, MP4 (in sampled files)
- Stored at: Google Cloud Storage, gs://zapbench-release (us)
- Access: public, readable without an account
- Category: Neuroscience
- Source: https://arxiv.org/abs/2503.02618
- Page: https://scigantic.com/archive/c6e91c7b-62b0-583b-9bf1-702ae912f940
- Keywords: Neuroscience, calcium imaging, electron microscopy, image segmentation, benchmark, PNG, MP4, Google Cloud Storage

## About

Data release of the Zebrafish Activity Prediction Benchmark (ZAPBench) from Google Research and HHMI Janelia, a benchmark for forecasting cellular-resolution neural activity across an entire vertebrate brain. Per the ICLR paper, the dataset contains 4D light-sheet microscopy calcium imaging recordings of over 70,000 neurons in a larval zebrafish brain, with motion-stabilized volumes and voxel-level cell segmentations; the same brain is being mapped at synaptic resolution by electron microscopy. Functional recordings were acquired in the Ahrens lab at Janelia, segmentation annotations by the CellMap Project Team, and alignment, segmentation (flood-filling network, FFN) and trace extraction by Google Research. Measured 2026-11-13: top-level folders volumes, dataframes, ffn_checkpoints, inference, neuroglancer, fluroglancer, notebooks and figures; volumes/20240930/ holds 28 products including raw, aligned and df_over_f (dF/F) activity volumes, anatomy stacks, annotations, segmentation, stimuli and traces. The traces array is Zarr v3, float32, shape 7879 x 71721 (timesteps x neurons), with a 1.98 GB traces.zip copy. Open the Zarr/TensorStore volumes with tensorstore or zarr in Python, view them in Neuroglancer, and follow the publisher's datasets colab. Datasets are licensed CC BY 4.0 per volumes/README.md; cite the ZAPBench paper (Lueckmann et al., ICLR 2025).

## Load the data

```python
import gcsfs
fs = gcsfs.GCSFileSystem(token='anon')
fs.ls('zapbench-release/', detail=False)[:10]
```

Lists the files with anonymous access. Runs on any machine, with no cloud account.

## Data structure

File types in sampled files: PNG, MP4

Example file: `assets/activity_fg.png`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/c6e91c7b-62b0-583b-9bf1-702ae912f940
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/c6e91c7b-62b0-583b-9bf1-702ae912f940
- Over MCP: `get_schema_card` with id `c6e91c7b-62b0-583b-9bf1-702ae912f940` at https://api.scigantic.com/mcp

## Related topics

- [Neuroscience datasets](https://scigantic.com/topics/neuroscience)
- [Benchmark datasets](https://scigantic.com/topics/benchmark)
- [Image segmentation datasets](https://scigantic.com/topics/image-segmentation)
- [PNG datasets](https://scigantic.com/topics/png)
- [Calcium imaging datasets](https://scigantic.com/topics/calcium-imaging)
- [Electron microscopy datasets](https://scigantic.com/topics/electron-microscopy)
- [MP4 datasets](https://scigantic.com/topics/mp4)
