{
  "annotation_id": "5d0aa6d7-4ecf-4a63-835a-74c928bb4060",
  "date_created": "2026-02-17T05:53:19.948000Z",
  "date_modified": "2026-02-17T05:53:19.948000Z",
  "file_hash": "f6b7a2054fef28f4f5b3ac5051fed503294780d60f00aa46e4b61938899ee09c",
  "private": false,
  "record": {
    "abstract": "This paper introduces a novel application of Video Joint-Embedding Predictive Architectures (V-JEPAs) for Facial Expression Recognition (FER). Departing from conventional pre-training methods for video understanding that rely on pixel-level reconstructions, V-JEPAs learn by predicting embeddings of masked regions from the embeddings of unmasked regions. This enables the trained encoder to not capture irrelevant information about a given video like the color of a region of pixels in the background. Using a pre-trained V-JEPA video encoder, we train shallow classifiers using the RAVDESS and CREMA-D datasets, achieving state-of-the-art performance on RAVDESS and outperforming all other vision-based methods on CREMA-D (+1.48 WAR). Furthermore, cross-dataset evaluations reveal strong generalization capabilities, demonstrating the potential of purely embedding-based pre-training approaches to advance FER. We release our code at https://github.com/lennarteingunia/vjepa-for-fer.",
    "arxiv_id": "2601.09524",
    "authors": [
      "Lennart Eing",
      "Cristina Luna-Jim\u00e9nez",
      "Silvan Mertes",
      "Elisabeth Andr\u00e9"
    ],
    "categories": [
      "cs.CV",
      "cs.HC"
    ],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "title": "Video Joint-Embedding Predictive Architectures for Facial Expression Recognition",
    "url": "https://arxiv.org/abs/2601.09524",
    "version": "v1"
  },
  "schema_id": "dorsal/arxiv",
  "source": {
    "execution_id": "e339c35c-3cac-4c05-953c-996719226c2c",
    "id": "arXiv Dataset",
    "type": "Model",
    "variant": "snapshot-2026-01-17",
    "version": "0.1.0"
  },
  "user_id": 1000002
}