Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,53 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
---
|
| 4 |
+
# [EMNLP'23] ReSee: Responding through Seeing Fine-grained Visual Knowledge in Open-domain Dialogue
|
| 5 |
+
|
| 6 |
+
ArXiv: https://arxiv.org/abs/2305.13602
|
| 7 |
+
|
| 8 |
+
Code: https://github.com/ImKeTT/ReSee
|
| 9 |
+
|
| 10 |
+
This is the processed data for ReSee, more raw image data are coming...
|
| 11 |
+
|
| 12 |
+
The data should looks like this:
|
| 13 |
+
```
|
| 14 |
+
.
|
| 15 |
+
├── ./processed_resee_data
|
| 16 |
+
├── dd # Contains proccessed entity-level image features and annotations of DailyDialogue
|
| 17 |
+
├── processed_img_features
|
| 18 |
+
└── img_clip_features.pt
|
| 19 |
+
├── test_v0.json
|
| 20 |
+
├── valid_v0.json
|
| 21 |
+
└── train_v0.json
|
| 22 |
+
├── wow # Contains proccessed entity-level image features and annotations of Wizard of Wikipedia
|
| 23 |
+
├── processed_img_features
|
| 24 |
+
└── img_clip_features.pt
|
| 25 |
+
├── test_random_v0.json
|
| 26 |
+
├── test_topic_v0.json
|
| 27 |
+
├── train_v0.json
|
| 28 |
+
├── valid_random_v0.json
|
| 29 |
+
└── valid_topic_v0.json
|
| 30 |
+
└── shared # Turn-level image features
|
| 31 |
+
├── coco
|
| 32 |
+
├── flickr30
|
| 33 |
+
├── nocaps
|
| 34 |
+
├── openimagev6
|
| 35 |
+
├── processed_img_features_clip_base # turn-level image features processed by ViT base
|
| 36 |
+
├── coco_train_clip_vis_fea.pt
|
| 37 |
+
├── coco_val_clip_vis_fea.pt
|
| 38 |
+
├── flickr30_clip_vis_fea.pt
|
| 39 |
+
├── nocaps_clip_vis_fea.pt
|
| 40 |
+
├── openimagev6_test_clip_vis_fea.pt
|
| 41 |
+
├── openimagev6_train_clip_vis_fea.pt
|
| 42 |
+
├── openimagev6_val_clip_vis_fea.pt
|
| 43 |
+
└── oodcv-counterfactual.json
|
| 44 |
+
└── processed_img_features_clip_large # turn-level image features processed by ViT large
|
| 45 |
+
├── coco_train_clip_vis_fea.pt
|
| 46 |
+
├── coco_val_clip_vis_fea.pt
|
| 47 |
+
├── flickr30_clip_vis_fea.pt
|
| 48 |
+
├── nocaps_clip_vis_fea.pt
|
| 49 |
+
├── openimagev6_test_clip_vis_fea.pt
|
| 50 |
+
├── openimagev6_train_clip_vis_fea.pt
|
| 51 |
+
├── openimagev6_val_clip_vis_fea.pt
|
| 52 |
+
└── oodcv-counterfactual.json
|
| 53 |
+
```
|