Update README.md
Browse files
README.md
CHANGED
|
@@ -50,8 +50,11 @@ The intention for this format is to teach a model to associate multiple modaliti
|
|
| 50 |
- **Text**: Not including the captions and transcripts, the “text” portion is a concatenation of Youtube’s original English transcripts associated with the above media of around 1–40 words per data record.
|
| 51 |
|
| 52 |
- Dataset Size:
|
| 53 |
-
- **About
|
| 54 |
-
- **About
|
|
|
|
|
|
|
|
|
|
| 55 |
|
| 56 |
## File Organization
|
| 57 |
- Each data entry follows the `<video><image(s)><audio><text>` structure as described above.
|
|
|
|
| 50 |
- **Text**: Not including the captions and transcripts, the “text” portion is a concatenation of Youtube’s original English transcripts associated with the above media of around 1–40 words per data record.
|
| 51 |
|
| 52 |
- Dataset Size:
|
| 53 |
+
- **About 7,000,000 records.**
|
| 54 |
+
- **About 15,000,000 images, each captioned with FLorence-2.**
|
| 55 |
+
- **About 30,000,000 audio snippets, about half of which transcribed with Whisper-large, and half with Youtube ASR.**
|
| 56 |
+
- **Divided into about 12K shards of about 600 records, each in a parquet file and a corresponding .tar.gz file for the media.**
|
| 57 |
+
- **About 14TB in total.***
|
| 58 |
|
| 59 |
## File Organization
|
| 60 |
- Each data entry follows the `<video><image(s)><audio><text>` structure as described above.
|