Research map / Video / Video generation
GePSAn: Generative Procedure Step Anticipation in Cooking Videos
Mohamed A. Abdelsalam, Samrudhdhi B. Rangrej, Isma Hadji, Nikita Dvornik, Konstantinos G. Derpanis, Afsaneh Fazly
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 |
|---|---|---|---|
| use a generative CVAE recipe encoder with five sampled predictions instead of a single prediction | GePSAn · YouCookII video, unseen split, finetuned | BLEU4: 4.8 -> 7.9 | Improved |
| replace baseline with GePSAn under multiple-prediction evaluation | baseline with nucleus sampling · YouCookII video, unseen split, finetuned | ingredient recall (ING): 27.8 -> 31.6 | Improved |
| add auxiliary MSE loss L_aux to train CVAE (ablation) | GePSAn · Recipe1M+ (text), multiple predictions | ingredient recall (ING): 29.4 -> 37.2 | Improved |
| add next-step prediction loss L_pred (ablation) | GePSAn · Recipe1M+ (text), multiple predictions | BLEU4: 7.7 -> 10.7 | Improved |
| add sentence reconstruction loss L_rec (ablation) | GePSAn · Recipe1M+ (text) | ingredient recall (ING), multiple: 36.6 -> 37.2 | No clear effect |
| keep the KL divergence term in the objective (ablation) | GePSAn · Recipe1M+ (text) | training stability | Improved |
| transfer text-pretrained model to video zero-shot via aligned UniVL encoders | GePSAn (M) · YouCookII video, unseen split | ingredient recall (ING): 30.0 zero-shot vs 27.8 finetuned baseline | Improved |
| finetune on video data after text pretraining | GePSAn (M) · YouCookII video, unseen split | BLEU4: 3.7 -> 7.9 | Improved |
| use frozen UniVL text encoder for text pretraining instead of learning from scratch | GePSAn · Recipe1M+ (text), single prediction | VERB recall: 29.4 -> 28.5 | No clear effect |
| train from scratch on YouCookII without text pretraining (ablation) | GePSAn · YouCookII | performance | Hurt |
| add cross-modal contrastive distillation to the baseline | baseline · YouCookII video, seen split, multiple | BLEU4: 5.9 -> 6.8 | Improved |
| zero-shot dataset transfer in single-prediction setting vs baseline | GePSAn (S) · YouCookII text, unseen split | ingredient recall (ING): 22.9 -> 20.0 | Hurt |
Closest papers by meaning
- VEDIT: Latent Prediction Architecture For Procedural Video Representation Learning (ICLR 2025 · similarity 0.89)
- Zero-Shot Anticipation for Instructional Activities (ICCV 2019 · similarity 0.89)
- Can't Make an Omelette Without Breaking Some Eggs: Plausible Action Anticipation Using Large Video-Language Models (CVPR 2024 · similarity 0.89)
- AntGPT: Can Large Language Models Help Long-term Action Anticipation from Videos? (ICLR 2024 · similarity 0.88)
- VG-TVP: Multimodal Procedural Planning via Visually Grounded Text-Video Prompting (AAAI 2025 · similarity 0.86)
- Towards Understanding Future: Consistency Guided Probabilistic Modeling for Action Anticipation (AAAI 2024 · similarity 0.86)
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.
