MovingFashion Retrieval

Frozen MTEB/MOEB representation of the official MovingFashion test split for video-to-shop-image fashion retrieval.

Construction

The dataset is derived from the official MovingFashion release and its test.json associations. The MTEB construction script audits both official splits against the source code at revision 4ca15d147ce87f0385c0c9779eac49e55c727ec8, checks that train and test share no product IDs or media paths, verifies every annotation reference against the archive, pins the known source omissions, and decodes all published test media. The source archive SHA-256 for this build is 20ae89a67a58d3dfc2304c2533d5a5c684eb6888aec9a7d7d3eda22b0a81f5f4.

Media paths are used as IDs. Repeated paths are collapsed without discarding associations, so the source's multi-positive relevance structure is preserved. The Hub configs use the standard MTEB representation: queries contains video queries, corpus contains shop images, and qrels contains binary relevance judgments.

The official archive omits 22 train videos and one annotated test video. The missing test query and its unusable qrel are excluded; its available shop image remains in the corpus as a distractor. The construction script pins and reports all 23 source omissions rather than silently dropping them.

Evaluation contents

• Video-to-image: 1328 queries, 1341 corpus images, and 1341 qrels.
• Corpus images without a qrel: 1.
• Source difficulty labels: 0 is hard and 1 is regular. They are audited during construction but are not used to filter the benchmark.

Source protocol and baselines

The source benchmark evaluates video-to-shop retrieval with top-k accuracy. The paper reports SEAM Match-RCNN top-1/5/10/20 accuracy of .49/.80/.89/.94 overall, .55/.86/.94/.97 on the regular subset, and .30/.62/.76/.87 on the hard subset. Those numbers use the original task-specific detector and are context rather than directly comparable guarantees for generic embedding models.

License, provenance, and limitations

The official repository labels the work CC BY-NC-SA 4.0 and says the dataset is available for academic purposes. This derived release therefore retains CC BY-NC-SA 4.0, attribution, non-commercial, and share-alike requirements. Source videos originated on Instagram and shop images on Net-A-Porter; underlying media, privacy, publicity, trademark, and platform rights may remain with their respective owners. The paper states that faces were blurred. Users remain responsible for determining whether their use complies with the license, source-platform terms, and applicable law.

Citation

@misc{godi2021movingfashion,
  title = {MovingFashion: a Benchmark for the Video-to-Shop Challenge},
  author = {Godi, Marco and Joppi, Christian and Skenderi, Geri and Cristani, Marco},
  year = {2021},
  eprint = {2110.02627},
  archivePrefix = {arXiv},
  primaryClass = {cs.CV},
}

See the paper and official code.