carlosj934 commited on
Commit
92b3aa8
·
1 Parent(s): 1efdfe3

Add 1 new BJJ samples - Version 1.2.0

Browse files
Files changed (4) hide show
  1. README.md +17 -36
  2. data-00000-of-00001.arrow +2 -2
  3. dataset_info.json +0 -4
  4. state.json +1 -1
README.md CHANGED
@@ -3,22 +3,20 @@ license: mit
3
  task_categories:
4
  - image-classification
5
  - keypoint-detection
6
- - object-detection
7
  tags:
8
  - martial-arts
9
  - bjj
10
  - brazilian-jiu-jitsu
11
  - pose-detection
12
  - sports-analysis
13
- - keypoint-detection
14
  - submissions
15
  - grappling
16
  - computer-vision
17
  language:
18
  - en
19
  size_categories:
20
- - 1K<n<10K
21
- version: 1.1.0
22
  ---
23
 
24
  # BJJ Positions & Submissions Dataset
@@ -34,7 +32,7 @@ This dataset contains pose keypoint annotations for Brazilian Jiu-Jitsu (BJJ) co
34
  - **Keypoint format**: MS-COCO (17 keypoints per person)
35
  - **Data format**: [x, y, confidence] for each keypoint
36
  - **Last updated**: 2025-07-21
37
- - **Version**: 1.1.0
38
 
39
  ### Supported Tasks
40
 
@@ -46,10 +44,10 @@ This dataset contains pose keypoint annotations for Brazilian Jiu-Jitsu (BJJ) co
46
 
47
  ## Recent Updates
48
 
49
- ### Version 1.1.0 (2025-07-21)
50
- - Added new samples across various positions
51
- - Improved position coverage and diversity
52
- - Enhanced submission technique annotations
53
 
54
  ### Position Distribution
55
 
@@ -72,25 +70,15 @@ This dataset contains pose keypoint annotations for Brazilian Jiu-Jitsu (BJJ) co
72
  - `total_keypoints`: Total visible keypoints across both athletes
73
  - `date_added`: Date when sample was added to dataset
74
 
75
- ### Position Classes
76
-
77
- The dataset includes the following BJJ positions and submissions:
78
-
79
- **Guard Positions:**
80
- - `open_guard1/2`: Open guard with athlete designation
81
- - `closed_guard1/2`: Closed guard with athlete designation
82
- - `half_guard1/2`: Half guard with athlete designation
83
- - `50_50_guard`: Equal leg entanglement position
84
 
85
- **Top Control Positions:**
86
- - `mount1/2`: Mount position with athlete designation
87
- - `side_control1/2`: Side control with athlete designation
88
- - `back1/2`: Back control with athlete designation
89
- - `turtle1/2`: Turtle position with athlete designation
90
 
91
- **Neutral/Transition:**
92
- - `standing`: Standing position
93
- - `takedown1/2`: Takedown attempt with initiator designation
94
 
95
  ## Usage
96
 
@@ -104,7 +92,7 @@ dataset = load_dataset("carlosj934/BJJ_Positions_Submissions")
104
  sample = dataset['train'][0]
105
  print(f"Position: {sample['position']}")
106
  print(f"Number of people: {sample['num_people']}")
107
- print(f"Athlete 1 keypoints: {sample['pose1_keypoints']}")
108
 
109
  # Filter by specific positions
110
  guard_samples = dataset['train'].filter(lambda x: 'guard' in x['position'])
@@ -122,13 +110,6 @@ The dataset is continuously updated with new BJJ position and submission samples
122
  - **Coverage**: 1/18+ positions represented
123
  - **Focus**: High-quality pose annotations for training robust BJJ classifiers
124
 
125
- ### Data Quality
126
-
127
- - All poses manually verified and labeled
128
- - Multiple camera angles per position
129
- - Diverse athlete body types and sizes
130
- - Clear, unobstructed pose visibility
131
-
132
  ## Applications
133
 
134
  This dataset can be used for:
@@ -148,7 +129,7 @@ If you use this dataset in your research, please cite:
148
  title={BJJ Positions and Submissions Dataset},
149
  author={Carlos J},
150
  year={2025},
151
- version={1.1.0},
152
  publisher={Hugging Face},
153
  url={https://huggingface.co/datasets/carlosj934/BJJ_Positions_Submissions}
154
  }
@@ -160,4 +141,4 @@ MIT License - See LICENSE file for details.
160
 
161
  ## Contact
162
 
163
- For questions or contributions, please reach out through the Hugging Face dataset page or open an issue in the associated repository.
 
3
  task_categories:
4
  - image-classification
5
  - keypoint-detection
 
6
  tags:
7
  - martial-arts
8
  - bjj
9
  - brazilian-jiu-jitsu
10
  - pose-detection
11
  - sports-analysis
 
12
  - submissions
13
  - grappling
14
  - computer-vision
15
  language:
16
  - en
17
  size_categories:
18
+ - n<1K
19
+ version: 1.2.0
20
  ---
21
 
22
  # BJJ Positions & Submissions Dataset
 
32
  - **Keypoint format**: MS-COCO (17 keypoints per person)
33
  - **Data format**: [x, y, confidence] for each keypoint
34
  - **Last updated**: 2025-07-21
35
+ - **Version**: 1.2.0
36
 
37
  ### Supported Tasks
38
 
 
44
 
45
  ## Recent Updates
46
 
47
+ ### Version 1.2.0 (2025-07-21)
48
+ - Added 1 total samples
49
+ - Improved data structure for better compatibility
50
+ - Enhanced position annotations
51
 
52
  ### Position Distribution
53
 
 
70
  - `total_keypoints`: Total visible keypoints across both athletes
71
  - `date_added`: Date when sample was added to dataset
72
 
73
+ ### Keypoint Format
 
 
 
 
 
 
 
 
74
 
75
+ Uses MS-COCO 17-keypoint format:
76
+ 0. nose, 1. left_eye, 2. right_eye, 3. left_ear, 4. right_ear
77
+ 5. left_shoulder, 6. right_shoulder, 7. left_elbow, 8. right_elbow
78
+ 9. left_wrist, 10. right_wrist, 11. left_hip, 12. right_hip
79
+ 13. left_knee, 14. right_knee, 15. left_ankle, 16. right_ankle
80
 
81
+ Each keypoint: [x, y, confidence] where confidence 0.0-1.0
 
 
82
 
83
  ## Usage
84
 
 
92
  sample = dataset['train'][0]
93
  print(f"Position: {sample['position']}")
94
  print(f"Number of people: {sample['num_people']}")
95
+ print(f"Athlete 1 keypoints: {len(sample['pose1_keypoints'])}")
96
 
97
  # Filter by specific positions
98
  guard_samples = dataset['train'].filter(lambda x: 'guard' in x['position'])
 
110
  - **Coverage**: 1/18+ positions represented
111
  - **Focus**: High-quality pose annotations for training robust BJJ classifiers
112
 
 
 
 
 
 
 
 
113
  ## Applications
114
 
115
  This dataset can be used for:
 
129
  title={BJJ Positions and Submissions Dataset},
130
  author={Carlos J},
131
  year={2025},
132
+ version={1.2.0},
133
  publisher={Hugging Face},
134
  url={https://huggingface.co/datasets/carlosj934/BJJ_Positions_Submissions}
135
  }
 
141
 
142
  ## Contact
143
 
144
+ For questions or contributions, please reach out through the Hugging Face dataset page.
data-00000-of-00001.arrow CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2666e3cd35002869b11070a9d67dac56c29673a1d1fc99ee89ca23da33dd1ce2
3
- size 3176
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5f60dad2bd584d70e0c118c8492d231825296a0547e2f08b7aca5c350552d861
3
+ size 3328
dataset_info.json CHANGED
@@ -24,10 +24,8 @@
24
  "dtype": "float32",
25
  "_type": "Value"
26
  },
27
- "length": 3,
28
  "_type": "List"
29
  },
30
- "length": 17,
31
  "_type": "List"
32
  },
33
  "pose1_num_keypoints": {
@@ -40,10 +38,8 @@
40
  "dtype": "float32",
41
  "_type": "Value"
42
  },
43
- "length": 3,
44
  "_type": "List"
45
  },
46
- "length": 17,
47
  "_type": "List"
48
  },
49
  "pose2_num_keypoints": {
 
24
  "dtype": "float32",
25
  "_type": "Value"
26
  },
 
27
  "_type": "List"
28
  },
 
29
  "_type": "List"
30
  },
31
  "pose1_num_keypoints": {
 
38
  "dtype": "float32",
39
  "_type": "Value"
40
  },
 
41
  "_type": "List"
42
  },
 
43
  "_type": "List"
44
  },
45
  "pose2_num_keypoints": {
state.json CHANGED
@@ -4,7 +4,7 @@
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
- "_fingerprint": "f00451cd30b9b0f1",
8
  "_format_columns": null,
9
  "_format_kwargs": {},
10
  "_format_type": null,
 
4
  "filename": "data-00000-of-00001.arrow"
5
  }
6
  ],
7
+ "_fingerprint": "67ff4e686df11486",
8
  "_format_columns": null,
9
  "_format_kwargs": {},
10
  "_format_type": null,