Hugging Face's logo

PulpBio
/
LUNA

PyTorch
English
luna
eeg
biosignal
time-series
cross-attention
foundation-model
self-supervised
masked-modeling
transformer
rope
neuroscience
neurips-2025
Eval Results (legacy)
Model card Files
xet
 Community
thorir commited on
Commit
4e23ccf
·
verified ·
1 Parent(s): 8bdf01e

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +205 -1
README.md CHANGED
@@ -2,13 +2,217 @@
2
  license: cc-by-nd-4.0
3
  language:
4
  - en
 
 
5
  tags:
6
  - eeg
 
7
  - time-series
8
  - cross-attention
9
  - foundation-model
 
 
 
 
10
  - neuroscience
11
- library_name: pytorch
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
12
  ---
13
 
14
 
 
2
  license: cc-by-nd-4.0
3
  language:
4
  - en
5
+ library_name: pytorch
6
+ pipeline_tag: feature-extraction
7
  tags:
8
  - eeg
9
+ - biosignal
10
  - time-series
11
  - cross-attention
12
  - foundation-model
13
+ - self-supervised
14
+ - masked-modeling
15
+ - transformer
16
+ - rope
17
  - neuroscience
18
+ - neurips-2025
19
+ datasets:
20
+ - TUEG
21
+ - Siena
22
+ - TUAB
23
+ - TUAR
24
+ - TUSL
25
+ - SEED-V
26
+ metrics:
27
+ - balanced_accuracy
28
+ - roc_auc
29
+ - pr_auc
30
+ - f1
31
+ - cohen_kappa
32
+ model-index:
33
+ - name: LUNA-Base
34
+ results:
35
+ - task:
36
+ type: time-series-classification
37
+ name: EEG Abnormality Detection
38
+ dataset:
39
+ type: TUAB
40
+ name: TUH EEG Abnormal Corpus (TUAB)
41
+ metrics:
42
+ - type: balanced_accuracy
43
+ value: 80.63
44
+ name: Balanced Accuracy (%)
45
+ - type: roc_auc
46
+ value: 0.8868
47
+ name: AUROC
48
+ - type: pr_auc
49
+ value: 0.8953
50
+ name: AUC-PR
51
+ - task:
52
+ type: time-series-classification
53
+ name: EEG Artifact Detection
54
+ dataset:
55
+ type: TUAR
56
+ name: TUH EEG Artifact Corpus (TUAR)
57
+ metrics:
58
+ - type: roc_auc
59
+ value: 0.902
60
+ name: AUROC
61
+ - type: pr_auc
62
+ value: 0.495
63
+ name: AUC-PR
64
+ - task:
65
+ type: time-series-classification
66
+ name: EEG Slowing Classification
67
+ dataset:
68
+ type: TUSL
69
+ name: TUH EEG Slowing Corpus (TUSL)
70
+ metrics:
71
+ - type: roc_auc
72
+ value: 0.767
73
+ name: AUROC
74
+ - type: pr_auc
75
+ value: 0.301
76
+ name: AUC-PR
77
+ - task:
78
+ type: time-series-classification
79
+ name: EEG Emotion Recognition
80
+ dataset:
81
+ type: SEED-V
82
+ name: SEED-V
83
+ metrics:
84
+ - type: balanced_accuracy
85
+ value: 37.30
86
+ name: Balanced Accuracy (%)
87
+ - type: cohen_kappa
88
+ value: 0.1831
89
+ name: Cohen's Kappa
90
+ - type: f1
91
+ value: 0.3389
92
+ name: Weighted F1
93
+ config: weighted
94
+ - name: LUNA-Large
95
+ results:
96
+ - task:
97
+ type: time-series-classification
98
+ name: EEG Abnormality Detection
99
+ dataset:
100
+ type: TUAB
101
+ name: TUH EEG Abnormal Corpus (TUAB)
102
+ metrics:
103
+ - type: balanced_accuracy
104
+ value: 80.96
105
+ name: Balanced Accuracy (%)
106
+ - type: roc_auc
107
+ value: 0.8924
108
+ name: AUROC
109
+ - type: pr_auc
110
+ value: 0.8986
111
+ name: AUC-PR
112
+ - task:
113
+ type: time-series-classification
114
+ name: EEG Artifact Detection
115
+ dataset:
116
+ type: TUAR
117
+ name: TUH EEG Artifact Corpus (TUAR)
118
+ metrics:
119
+ - type: roc_auc
120
+ value: 0.918
121
+ name: AUROC
122
+ - type: pr_auc
123
+ value: 0.505
124
+ name: AUC-PR
125
+ - task:
126
+ type: time-series-classification
127
+ name: EEG Slowing Classification
128
+ dataset:
129
+ type: TUSL
130
+ name: TUH EEG Slowing Corpus (TUSL)
131
+ metrics:
132
+ - type: roc_auc
133
+ value: 0.771
134
+ name: AUROC
135
+ - type: pr_auc
136
+ value: 0.293
137
+ name: AUC-PR
138
+ - task:
139
+ type: time-series-classification
140
+ name: EEG Emotion Recognition
141
+ dataset:
142
+ type: SEED-V
143
+ name: SEED-V
144
+ metrics:
145
+ - type: balanced_accuracy
146
+ value: 39.18
147
+ name: Balanced Accuracy (%)
148
+ - type: cohen_kappa
149
+ value: 0.2073
150
+ name: Cohen's Kappa
151
+ - type: f1
152
+ value: 0.3586
153
+ name: Weighted F1
154
+ config: weighted
155
+ - name: LUNA-Huge
156
+ results:
157
+ - task:
158
+ type: time-series-classification
159
+ name: EEG Abnormality Detection
160
+ dataset:
161
+ type: TUAB
162
+ name: TUH EEG Abnormal Corpus (TUAB)
163
+ metrics:
164
+ - type: balanced_accuracy
165
+ value: 81.57
166
+ name: Balanced Accuracy (%)
167
+ - type: roc_auc
168
+ value: 0.8957
169
+ name: AUROC
170
+ - type: pr_auc
171
+ value: 0.9029
172
+ name: AUC-PR
173
+ - task:
174
+ type: time-series-classification
175
+ name: EEG Artifact Detection
176
+ dataset:
177
+ type: TUAR
178
+ name: TUH EEG Artifact Corpus (TUAR)
179
+ metrics:
180
+ - type: roc_auc
181
+ value: 0.921
182
+ name: AUROC
183
+ - type: pr_auc
184
+ value: 0.528
185
+ name: AUC-PR
186
+ - task:
187
+ type: time-series-classification
188
+ name: EEG Slowing Classification
189
+ dataset:
190
+ type: TUSL
191
+ name: TUH EEG Slowing Corpus (TUSL)
192
+ metrics:
193
+ - type: roc_auc
194
+ value: 0.802
195
+ name: AUROC
196
+ - type: pr_auc
197
+ value: 0.289
198
+ name: AUC-PR
199
+ - task:
200
+ type: time-series-classification
201
+ name: EEG Emotion Recognition
202
+ dataset:
203
+ type: SEED-V
204
+ name: SEED-V
205
+ metrics:
206
+ - type: balanced_accuracy
207
+ value: 39.00
208
+ name: Balanced Accuracy (%)
209
+ - type: cohen_kappa
210
+ value: 0.2037
211
+ name: Cohen's Kappa
212
+ - type: f1
213
+ value: 0.3506
214
+ name: Weighted F1
215
+ config: weighted
216
  ---
217
 
218