Update README.md
#3
by
Harini
- opened
README.md
CHANGED
@@ -58,3 +58,39 @@ configs:
|
|
58 |
- split: test
|
59 |
path: data/test-*
|
60 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
58 |
- split: test
|
59 |
path: data/test-*
|
60 |
---
|
61 |
+
|
62 |
+
|
63 |
+
## Dataset Description
|
64 |
+
|
65 |
+
- **Website:** https://behavior-in-the-wild.github.io/memorability.html
|
66 |
+
- **Paper:** https://arxiv.org/pdf/2309.00378
|
67 |
+
|
68 |
+
### Dataset Summary
|
69 |
+
LAMDBA is a long term ad memorability dataset, featuring data from 1749 participants and 2205 ads across 276 brands.
|
70 |
+
|
71 |
+
## Dataset Structure
|
72 |
+
|
73 |
+
```python
|
74 |
+
from datasets import load_dataset
|
75 |
+
ds = load_dataset("behavior-in-the-wild/LAMBDA")
|
76 |
+
ds
|
77 |
+
|
78 |
+
DatasetDict({
|
79 |
+
train: Dataset({
|
80 |
+
features: ['video_id', 'recall_score', 'youtube_id', 'ad_details'],
|
81 |
+
num_rows: 1964
|
82 |
+
})
|
83 |
+
test: Dataset({
|
84 |
+
features: ['video_id', 'recall_score', 'youtube_id', 'ad_details'],
|
85 |
+
num_rows: 219
|
86 |
+
})
|
87 |
+
})
|
88 |
+
```
|
89 |
+
|
90 |
+
### Data Fields
|
91 |
+
|
92 |
+
- `video_id`: identifier for the data sample
|
93 |
+
- `recall_score`: memorability score for the video between 0 to 1
|
94 |
+
- `youtube_id`: youtube id for the video
|
95 |
+
- `ad_details`: scene by scene features for each video
|
96 |
+
|