lyx97 commited on
Commit
d23d5db
1 Parent(s): 1a285e2

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +37 -0
README.md CHANGED
@@ -1,3 +1,40 @@
1
  ---
2
  license: cc-by-4.0
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: cc-by-4.0
3
  ---
4
+ # FETV
5
+
6
+ **FETV** is a benchmark for **F**ine-grained **E**valuation of open-domain **T**ext-to-**V**ideo generation
7
+
8
+ ## Overview
9
+ FETV consist of a diverse set of text prompts, categorized based on three orthogonal aspects: major content, attribute control, and prompt complexity.
10
+ ![](./Figures/categorization.png)
11
+
12
+ ## Data Statistics
13
+ FETV contains 619 text prompts. The data distributions over different categories are as follows
14
+ ![](./Figures/content_attribute_statistics.png)
15
+ ![](./Figures/complexity_statistics.png)
16
+
17
+ ## Data Format
18
+ All FETV data are all available in the file `fetv_data.json`. Each line is a data instance, which is formatted as:
19
+ ```
20
+ {
21
+ "video_id": 1006807024,
22
+ "prompt": "A mountain stream",
23
+ "major content": {
24
+ "spatial": ["scenery & natural objects"],
25
+ "temporal": ["fluid motions"]
26
+ },
27
+ "attribute control": {
28
+ "spatial": null,
29
+ "temporal": null
30
+ },
31
+ "prompt complexity": ["simple"],
32
+ "source": "WebVid",
33
+ "video_url": "https://ak.picdn.net/shutterstock/videos/1006807024/preview/stock-footage-a-mountain-stream.mp4"
34
+ }
35
+ ```
36
+ * "video_id" is the video identifier in the original dataset where the prompt comes from.
37
+ * "prompt" is the text prompt for text-to-video generation.
38
+ * "major content", "attribute control" and "prompt complexity" are the three orthogonal aspects for categorization.
39
+ * "source" denotes the original dataset where the prompt comes from, which can be "WebVid", "MSRVTT" or "ours".
40
+ * "video_url" is the url link of the reference video.