Dense Captioning 벤치마크
Dense Captioning on Visual Genome
mAP
- 2015-11-24 — FCLN: mAP 5.4
- 2019-04-02 — CAG-Net: mAP 10.5
- 2022-12-01 — GRiT (ViT-B): mAP 15.5
- 2024-01-31 — ControlCap: mAP 18.2
| Rank | Model | mAP | Paper | Code | Year |
|---|---|---|---|---|---|
| 1 | ControlCap | 18.2 | ControlCap: Controllable Region-level Captioning | callsys/controlcap | 2024 |
| 2 | GRiT (ViT-B) | 15.5 | GRiT: A Generative Region-to-text Transformer for Object Understanding | JialianW/GRiT | 2022 |
| 3 | CAG-Net | 10.5 | Context and Attribute Grounded Dense Captioning | 2019 | |
| 4 | FCLN | 5.4 | DenseCap: Fully Convolutional Localization Networks for Dense Captioning | jcjohnson/densecap | 2015 |