Phrase Grounding on Flickr30k entities (val)
87.1Recall@1FIBER-B
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| FIBER-BImage Backbone=Swin-B, #Pretrain Images (fine-grained)=860k2022.06 | 87.1 | 96.1 | 97.4 | — | |
| FIBERZero-shot=false2023.06 | 87.1 | 96.1 | 97.4 | — | |
| GLIP-LData=FourODs, GoldG, Cap24M2021.12 | 86.7 | 96.4 | 97.9 | — | |
| GLIPImage Backbone=Swin-L, #Pretrain Images (fine-grained)=27M2022.06 | 86.7 | 96.4 | 97.9 | — | |
| GLIPZero-shot=false2023.06 | 86.7 | 96.4 | 97.9 | — | |
| FIBER-B (w/o C.G. VLP)Image Backbone=Swin-B, #Pretrain Images (fine-grained)=860k, Note=without coarse-grained VL pretraining2022.06 | 86.2 | 96 | 97.6 | — | |
| GLIP-TData=O365, GoldG, Cap4M2021.12 | 85.7 | 95.4 | 96.9 | — | |
| GLIPImage Backbone=Swin-B, #Pretrain Images (fine-grained)=860k2022.06 | 85.7 | 95 | 96.2 | — | |
| GLIP-TData=O365, GoldG2021.12 | 84.8 | 94.9 | 96.3 | — | |
| GLIP-TData=GoldG2021.12 | 84 | 95.1 | 96.8 | — | |
| MDETR-ENB5†*Evaluation Protocol=ANY-BOX-PROTOCOL, Backbone=EfficientNet-B5, Pre-training=COCO, VG, Flickr 30k2021.04 | 83.6 | 93.4 | 95.1 | — | |
| MDETR-ENB5Data=GoldG+2021.12 | 83.6 | 93.4 | 95.1 | — | |
| MDETRImage Backbone=EN-B5, #Pretrain Images (fine-grained)=200k2022.06 | 83.6 | 93.4 | 95.1 | — | |
| MDETRZero-shot=false2023.06 | 83.6 | 93.4 | 95.1 | — | |
| MDETR-ENB3†*Evaluation Protocol=ANY-BOX-PROTOCOL, Backbone=EfficientNet-B3, Pre-training=COCO, VG, Flickr 30k2021.04 | 82.9 | 93.2 | 95.2 | — | |
| MDETR-R101†*Evaluation Protocol=ANY-BOX-PROTOCOL, Backbone=ResNet-101, Pre-training=COCO, VG, Flickr 30k2021.04 | 82.5 | 92.9 | 94.9 | — | |
| MDETR-RN101Data=GoldG+2021.12 | 82.5 | 92.9 | 94.9 | — | |
| MDETR-R101†*Evaluation Protocol=MERGED-BOXES-PROTOCOL, Backbone=ResNet-101, Pre-training=COCO, VG, Flickr 30k2021.04 | 82.3 | 91.8 | 93.7 | — | |
| MDETR-R101Evaluation Protocol=MERGED-BOXES-PROTOCOL, Backbone=ResNet-101, Pre-training=None2021.04 | 79 | 86.7 | 88.6 | — | |
| MDETR-R101Evaluation Protocol=ANY-BOX-PROTOCOL, Backbone=ResNet-101, Pre-training=None2021.04 | 78.9 | 88.8 | 90.8 | — | |
| KOSMOS-2Zero-shot=true2023.06 | 77.8 | 79.2 | 79.3 | — | |
| DDPNEvaluation Protocol=MERGED-BOXES-PROTOCOL, Backbone=ResNet-1012021.04 | 72.8 | — | — | — | |
| VisualBert+Evaluation Protocol=ANY-BOX-PROTOCOL, Backbone=ResNet-1012021.04 | 70.4 | 84.5 | 86.3 | — | |
| Visual-BERTImage Backbone=ResNet-101, #Pretrain Images (fine-grained)=120k2022.06 | 70.4 | 84.5 | 86.3 | — | |
| VisualBertZero-shot=false2023.06 | 70.4 | 84.5 | 86.3 | — | |
| VisualBertEvaluation Protocol=ANY-BOX-PROTOCOL, Backbone=ResNet-1012021.04 | 68.1 | 84 | 86.2 | — | |
| Ferret-13B2024.04 | — | — | — | 81.13 | |
| Ferret-7B2024.04 | — | — | — | 80.39 | |
| Ferret-v2-13B2024.04 | — | — | — | 85.33 | |
| Ferret-v2-7B2024.04 | — | — | — | 85.52 | |
| LLaVA-G2024.04 | — | — | — | 83.03 | |
| MDETRSpecifically fine-tuned in the second stage=true2024.04 | — | — | — | 82.3 | |
| Shikra-13B2024.04 | — | — | — | 77.41 | |
| Shikra-7B2024.04 | — | — | — | 75.84 | |
| UniTAB2024.04 | — | — | — | 78.76 |