Add dataset card with paper and code links
#1
by nielsr HF Staff - opened
README.md
CHANGED
|
@@ -1,3 +1,13 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: mit
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
task_categories:
|
| 4 |
+
- other
|
| 5 |
+
---
|
| 6 |
+
|
| 7 |
+
# ViRe: Medical Time Series Benchmarks
|
| 8 |
+
|
| 9 |
+
This repository contains the datasets/benchmarks used in the paper [Revitalizing Medical Time Series with Vision-Informed Retrieval: A Vision-Language Perspective](https://huggingface.co/papers/2609.34652).
|
| 10 |
+
|
| 11 |
+
The collection includes six subject-independent medical time series classification benchmarks: APAVA, ADFTD, TDBrain (EEG) and PTB, PTB-XL, MIMIC (ECG). They are used for classification tasks such as Alzheimer's disease, Parkinson's disease, and heart disease/myocardial infarction.
|
| 12 |
+
|
| 13 |
+
Code, training scripts, preprocessing notebooks, and reproducibility materials are available at: [https://github.com/Levi-Ackman/ViRe](https://github.com/Levi-Ackman/ViRe). For detailed instructions on data preprocessing, feature extraction, and training, please refer to the GitHub repository.
|