SciVid Fly-vs-Fly: Drosophila Social Behavior Video Clips for Action Classification (Google DeepMind SciVid Benchmark)
- Provider
- Google DeepMind (SciVid); source data Eyjolfsdottir et al., Caltech and HHMI
- License
- CC-BY-4.0 (source)
- Stored at
- Google Cloud Storage, gs://scivid/full/fly_vs_fly (us)
- Access
- Free and public. No account with the data provider is needed.
- Category
- Neuroscience
- Source
- https://arxiv.org/abs/2507.03578
About this dataset
Preprocessed Fly-vs-Fly videos of interacting fruit flies (Drosophila melanogaster) for frame-level social behavior classification, as packaged in the SciVid benchmark. Fly-vs-Fly (Eyjolfsdottir et al., Detecting social actions of fruit flies, ECCV 2014; Caltech and HHMI) has 22 hours of video of pairs of fruit flies engaged in 10 expert-annotated social interactions in three contexts: Boy meets boy, Aggression (male-male aggression under genetic manipulation) and Courtship. SciVid filtered the clips as described by Task Programming (Sun et al., CVPR 2021) and provides about 1 million training samples, with train, validation and test splits (for example test_fly_split_ids_shuffle.array_record-00000-of-00100) of short video windows and behavior labels. Used to test whether general video models transfer to animal behavior analysis in neuroscience and ethology. SciVid (Hasson et al., SciVid: Cross-Domain Evaluation of Video Models in Scientific Applications, ICCV 2025, arXiv:2507.03578) is a five-task benchmark for video foundation models in science; this folder is one task. Files are ArrayRecord shards, read with the array_record Python package or the loaders in the google-deepmind/scivid repository. License: SciVid materials CC BY 4.0 (Google DeepMind); the source Fly-vs-Fly data are CC0 at CaltechDATA (DOI 10.22002/D1.1893). Cite SciVid and Eyjolfsdottir et al. 2014.
How to access this dataset in Python
import gcsfs
fs = gcsfs.GCSFileSystem(token='anon')
fs.ls('scivid/full/fly_vs_fly/', detail=False)[:10]Lists the files with anonymous access. Runs on any machine, with no cloud account.
Data structure
Files
Example file: full/fly_vs_fly/
Analyze this dataset without downloading it
On Scigantic, SciVid Fly-vs-Fly: Drosophila Social Behavior Video Clips for Action Classification (Google DeepMind SciVid Benchmark) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.
This page is also available as Markdown and as JSON.