Lax3n commited on
Commit
bc9c567
·
verified ·
1 Parent(s): c0eb7b1

Publish a curated public release capped at 1 TB

Browse files
Files changed (3) hide show
  1. README.md +16 -9
  2. captions.parquet +2 -2
  3. selection.json +28 -0
README.md CHANGED
@@ -26,16 +26,19 @@ configs:
26
  [![Powered by Tornado API](https://tornadoapi.io/badge.svg)](https://tornadoapi.io?ref=hf-activitynet-badge)
27
 
28
  **15,941 videos** (391.3 GB) with 1,550 subtitle files, downloaded at source
29
- quality and re-hosted for direct use — no more dead YouTube links, no more flaky
30
- downloader scripts.
31
 
32
- > **Coverage:** 15,941 of the 19,994 source video IDs (79.7%). 4,053 source videos were unavailable at fetch time (private, removed, members-only or geo-blocked) and are excluded. The dataset is refreshed as more videos are delivered.
 
 
 
 
33
 
34
  ## What's inside
35
 
36
  - **`metadata.jsonl`** — one row per video: `video_id`, `video_url` (direct HTTPS download, CDN-backed), `subtitle_url`, `youtube_url`, `size_bytes`
37
  - The 15,941 MP4s (391.3 GB) and 1,550 subtitle files are hosted on Hugging Face storage and served via the `video_url` column
38
- - **`captions.parquet`** — 23,064 academic annotations (source: `annotations.csv`)
39
 
40
  ## Quick start
41
 
@@ -51,8 +54,10 @@ import requests
51
  r = requests.get(ds[0]["video_url"])
52
  open(f"{ds[0]['video_id']}.mp4", "wb").write(r.content)
53
 
54
- # academic annotations
 
55
  ann = load_dataset("TornadoLabs/activitynet", "annotations", split="train")
 
56
  ```
57
 
58
  ## How this dataset was built
@@ -75,8 +80,10 @@ annotations belong to the original dataset authors — cite their paper if you u
75
  them. This re-hosting is provided for reproducible research; if you are a rights
76
  holder and want content removed, open a discussion on this repo.
77
 
78
- ## Build your own video dataset
79
 
80
- Need a custom corpus (different IDs, other platforms, higher volume)?
81
- **[Tornado API](https://tornadoapi.io?ref=hf-activitynet-cta)** downloads YouTube at scale — bulk
82
- endpoints, S3/R2/HF delivery, subtitles and metadata included.
 
 
 
26
  [![Powered by Tornado API](https://tornadoapi.io/badge.svg)](https://tornadoapi.io?ref=hf-activitynet-badge)
27
 
28
  **15,941 videos** (391.3 GB) with 1,550 subtitle files, downloaded at source
29
+ quality and hosted for direct download.
 
30
 
31
+ > **Coverage:** 15,941 of the 19,994 source video IDs (79.7%). 4,053 source videos were unavailable at fetch time (private, removed, members-only or geo-blocked) and are excluded.
32
+
33
+
34
+ > **Public release:** capped at 1 TB (1,000,000,000,000 bytes), including media and indexes.
35
+ > This release retains 15,941 of the 15,941 previously hosted source videos. See `selection.json` for the selection method and byte counts. This is a selected subset, not the full upstream corpus.
36
 
37
  ## What's inside
38
 
39
  - **`metadata.jsonl`** — one row per video: `video_id`, `video_url` (direct HTTPS download, CDN-backed), `subtitle_url`, `youtube_url`, `size_bytes`
40
  - The 15,941 MP4s (391.3 GB) and 1,550 subtitle files are hosted on Hugging Face storage and served via the `video_url` column
41
+ - **`captions.parquet`** — 18,344 academic annotations (source: `annotations.csv`)
42
 
43
  ## Quick start
44
 
 
54
  r = requests.get(ds[0]["video_url"])
55
  open(f"{ds[0]['video_id']}.mp4", "wb").write(r.content)
56
 
57
+
58
+ # annotations for the published source videos
59
  ann = load_dataset("TornadoLabs/activitynet", "annotations", split="train")
60
+
61
  ```
62
 
63
  ## How this dataset was built
 
80
  them. This re-hosting is provided for reproducible research; if you are a rights
81
  holder and want content removed, open a discussion on this repo.
82
 
83
+ ## Request a larger or custom delivery
84
 
85
+ For larger volumes, different source IDs or a custom selection,
86
+ **[request a delivery quote from Tornado API](https://tornadoapi.io?ref=hf-activitynet-cta)**.
87
+ Deliverable coverage depends on source availability and applicable rights.
88
+ This is not a guarantee of the complete original corpus or a grant of commercial
89
+ rights to the source videos. The upstream terms continue to apply.
captions.parquet CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b2a231e9b6bdc103c46de1449cf5be0332910d3e4700ab582f5b7ae4107e027c
3
- size 1050182
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd95642a194794ede813c835093b0c9ffc6f93818cf33461ae1d2751eb7f0749
3
+ size 836661
selection.json ADDED
@@ -0,0 +1,28 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "version": 1,
3
+ "dataset": "activitynet",
4
+ "bucket_id": "TornadoLabs/staging",
5
+ "inventory_observed_at": "2026-09-29T16:25:32.838092+00:00",
6
+ "cap_bytes": 1000000000000,
7
+ "reserve_bytes": 1000000000,
8
+ "seed": "tornadolabs-public-v1",
9
+ "method": "sha256-weighted-round-robin-by-source-count",
10
+ "stratum_field": "unstratified",
11
+ "before_bytes": 391294886506,
12
+ "after_bucket_bytes": 391294886506,
13
+ "remove_bytes": 0,
14
+ "before_videos": 15941,
15
+ "after_videos": 15941,
16
+ "retained_unrecognized_objects": 0,
17
+ "strata": {
18
+ "unknown": {
19
+ "before_videos": 15941,
20
+ "after_videos": 15941,
21
+ "before_bytes": 391294886506,
22
+ "after_bytes": 391294886506
23
+ }
24
+ },
25
+ "plan_sha256": "52f572ccbe22c07bbeafe6fa9de6f6c68644c66fcf352e8aa1f4aa4ae119aca7",
26
+ "size_unit": "decimal bytes; 1 TB = 1,000,000,000,000 bytes",
27
+ "sampling_limitations": "Selected from previously hosted source videos, not the full upstream corpus. Weighted round robin follows source counts within the available annotation strata; the byte constraint can favor smaller groups near the cap. No new language or duration balancing."
28
+ }