Research map / Architectures and efficiency / RNNs and normalization
DeeCap: Dynamic Early Exiting for Efficient Image Captioning
Zhengcong Fei, Xu Yan, Shuhui Wang, Qi Tian
Check 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 imitated deep representations to the early-exit model (ablation) | DeeCap · MS COCO offline test, 4× speed-up | CIDEr: 127.8 -> 129.0 | Improved |
| add imitated deep representations to the early-exit model (ablation) | DeeCap · MS COCO offline test, 4× speed-up | FCS: 79.55 -> 82.40 | Improved |
| add imitated deep representations to the early-exit model (ablation) | DeeCap · MS COCO offline test, 2× speed-up | CIDEr: 128.3 -> 129.5 | Improved |
| use concatenation fusion instead of average fusion (ablation) | DeeCap · MS COCO offline test | BLEU-4: 38.3 -> 38.7 | Improved |
| replace concatenation fusion with attention-pooling (ablation) | DeeCap · MS COCO offline test | BLEU-4: 38.7 -> 38.6 | Hurt |
| replace concatenation fusion with SeqNN (ablation) | DeeCap · MS COCO offline test | BLEU-4: 38.7 -> 38.5 | Hurt |
| use dynamic early exiting with multilevel feature fusion instead of vanilla early exiting | image captioning on MS COCO Karpathy test · MS COCO Karpathy test | CIDEr: 126.3 -> 129.0 | Improved |
| use dynamic early exiting with multilevel feature fusion instead of vanilla early exiting | image captioning on MS COCO Karpathy test · MS COCO Karpathy test | SpeedUp: 4.54× -> 4.35× | Mixed |
Later papers that compared with this method
1 later accepted paper reports DeeCap as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- You Need Multiple Exiting: Dynamic Early Exiting for Accelerating Unified Vision Language Model (CVPR 2023 · better on 2 of 2 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
- SmallCap: Lightweight Image Captioning Prompted With Retrieval Augmentation (CVPR 2023 · similarity 0.87)
- Stack-Captioning: Coarse-to-Fine Learning for Image Captioning (AAAI 2018 · similarity 0.87)
- Cross-Domain Image Captioning With Discriminative Finetuning (CVPR 2023 · similarity 0.86)
- CapRL: Stimulating Dense Image Caption Capabilities via Reinforcement Learning (ICLR 2026 · similarity 0.85)
- CapsFusion: Rethinking Image-Text Data at Scale (CVPR 2024 · similarity 0.85)
- ViPCap: Retrieval Text-Based Visual Prompts for Lightweight Image Captioning (AAAI 2025 · similarity 0.84)
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.
