Research map / Video / Action recognition
VideoBERT: A Joint Model for Video and Language Representation Learning
Chen Sun, Austin Myers, Carl Vondrick, Kevin Murphy, Cordelia Schmid
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add video inputs (cross modal) to language prior (ablation) | VideoBERT · YouCook II action classification | verb top-5 (%): 6.9 -> 43.3 | Improved |
| add video inputs (cross modal) to language prior (ablation) | VideoBERT · YouCook II action classification | object top-5 (%): 15.3 -> 33.7 | Improved |
| fine-tune BERT on cooking videos language (ablation) | BERT language prior · YouCook II action classification | object top-1 (%): 0.0 -> 7.7 | Improved |
| increase pretraining videos from 10K to 300K (ablation) | VideoBERT · YouCook II action classification (pretraining data size) | verb top-5 (%): 15.5 -> 43.3 | Improved |
| increase pretraining videos from 10K to 300K (ablation) | VideoBERT · YouCook II action classification (pretraining data size) | object top-5 (%): 17.8 -> 33.7 | Improved |
| increase pretraining videos from 10K to 300K (ablation) | VideoBERT · YouCook II action classification (pretraining data size) | object top-1 (%): 2.9 -> 13.1 | Improved |
| add text via cross-modal pretraining instead of video-only (ablation) | VideoBERT feature extractor · YouCook II video captioning (val) | CIDEr: 0.47 -> 0.49 | Improved |
| add text via cross-modal pretraining instead of video-only (ablation) | VideoBERT feature extractor · YouCook II video captioning (val) | BLEU-4: 3.81 -> 4.04 | Improved |
| use VideoBERT features instead of S3D features | captioning transformer · YouCook II video captioning (val) | CIDEr: 0.31 -> 0.49 | Improved |
| concatenate S3D features with VideoBERT features (ablation) | captioning transformer · YouCook II video captioning (val) | CIDEr: 0.49 -> 0.55 | Improved |
| use VideoBERT cross modal instead of supervised S3D classifier | zero-shot classification · YouCook II action classification | verb top-1 (%): 16.1 -> 3.2 | Hurt |
| use VideoBERT cross modal instead of supervised S3D classifier | zero-shot classification · YouCook II action classification | verb top-5 (%): 46.9 -> 43.3 | No clear effect |
Later papers that compared with this method
3 later accepted papers report VideoBERT as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- LV-MAE: Learning Long Video Representations through Masked-Embedding Autoencoders (ICCV 2025 · better on 1 of 1 shared result)
- MECD+: Unlocking Event-Level Causal Graph Discovery for Video Reasoning (TPAMI 2025 · better on 0 of 1 shared result)
- Towards Long-Form Video Understanding (CVPR 2021 · better on 8 of 9 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- ActBERT: Learning Global-Local Video-Text Representations (CVPR 2020 · similarity 0.90)
- Less Is More: ClipBERT for Video-and-Language Learning via Sparse Sampling (CVPR 2021 · similarity 0.90)
- SwinBERT: End-to-End Transformers With Sparse Attention for Video Captioning (CVPR 2022 · similarity 0.89)
- Language Models with Image Descriptors are Strong Few-Shot Video-Language Learners (NeurIPS 2022 · similarity 0.89)
- ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks (NeurIPS 2019 · similarity 0.88)
- Distilling Vision-Language Models on Millions of Videos (CVPR 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
