File size: 3,113 Bytes
8606296
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
---
language:
- en
license: agpl-3.0
tags:
- video
- soccernet
- football
- soccer
- sport
- action-anticipation
size_categories: # Number of clips
- 10K<n<100K
source_datasets:
- SoccerNet/SN-BAS-2024
# HuggingFace does not have a task category for action anticipation. Therefore, I had to remove the task_categories tag
#task_categories:
#- action-anticipation
---

<!-- Builds from the template https://github.com/huggingface/huggingface_hub/blob/main/src/huggingface_hub/templates/datasetcard_template.md?plain=1 -->

# SoccerNet Challenge 2026 - Action Anticipation Dataset

<!-- Provide a quick summary of the dataset. -->

The SoccerNet Action Anticipation dataset splits the [2024 ball action spotting dataset](https://huggingface.co/datasets/SoccerNet/SN-BAS-2024) into 30 second clips, which can then be used to anticipate between 10 action classes that will happen 5 seconds into the future. This is the dataset used for 2026 SoccerNet Action Anticipation challenge.

## Relevant Links

<!-- Provide the basic links for the dataset. -->

- **Repository:** https://github.com/MohamadDalal/FAANTRA
- **Paper:** https://huggingface.co/papers/2504.12021

## Uses

<!-- Address questions around how the dataset is intended to be used. -->

### Direct Use

<!-- This section describes suitable use cases for the dataset. -->

The direct use of this dataset is to participate in the 2026 SoccerNet Action Anticipation challenge. For a quick demo the FAANTRA repository can be used to train and evaluate on the dataset.

### Other Use

The dataset can be used for further action anticipation research within the soccer field.

## Download

The setup_dataset_BAA.py script inside the FAANTRA repository can be used to download and setup the dataset.

## Dataset Structure

<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->

The dataset is structured as:
```
224p
|_train.zip
|_valid.zip
|_test.zip
|_challenge.zip
720p
|_train.zip
|_valid.zip
|_test.zip
|_challenge.zip
```
Each zip file contains a split at a specific resolution. Each split is structured as:
```
split
|_clip_1
  |_{224p|720p}.mp4
|_clip_2
  |_{224p|720p}.mp4
...
|_Labels-ball.json
```

The challenge split however, does not contain annotations, and therefore does not have the Labels-ball.json file.

## Citation

<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->

**BibTeX:**
```
@InProceedings{Dalal_2025_CVPR,
    author    = {Dalal, Mohamad and Xarles, Artur and Cioppa, Anthony and Giancola, Silvio and Van Droogenbroeck, Marc and Ghanem, Bernard and Clap\'es, Albert and Escalera, Sergio and Moeslund, Thomas B.},
    title     = {Action Anticipation from SoccerNet Football Video Broadcasts},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
    month     = {June},
    year      = {2025},
    pages     = {6126-6137}
}
```