Sign in to view source links and access this dataset
Description
MoViRec is a benchmark dataset for evaluating small video variational autoencoders. It contains 500 video patches, each 5 seconds long and 256x256 pixels, sourced from various AI video subreddits. The dataset was created by 'madebyollin' and was last updated on December 17, 2025.
Use Cases
Benchmarking video reconstruction quality based on the provided 500 video patches.
Evaluating feature reconstruction error using the included V-JEPA and JEDi SSv2 fine-tuned pooling block featurizer.
Testing video variational autoencoder performance via the provided evaluation notebook.
Strengths
Contains 500 video patches, providing a defined scale for testing.
Each video patch is 5 seconds long and 256x256 pixels, offering consistent input dimensions.
Includes a specific evaluation notebook and a choice of video featurizer for standardized testing.
Limitations
Row count for the full dataset is unknown, which may limit suitability assessment.
Column-level documentation is absent; field semantics must be inferred after download.
Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
Source
Video patches drawn from various AI video subreddits.
Collection Method
Patches were likely extracted or sampled from videos posted on subreddits.
Time Range
null
Freshness
Last updated 2025-12-17 02:15:05; freshness should be verified.
Geography
null
License is unknown; restrictions should be verified before use.