README: 2 datasets
Browse files
README.md
CHANGED
|
@@ -24,11 +24,23 @@ configs:
|
|
| 24 |
data_files:
|
| 25 |
- split: train
|
| 26 |
path: sidecar/IDRiD/*.parquet
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 27 |
---
|
| 28 |
|
| 29 |
# OphGen — curated public ophthalmic datasets
|
| 30 |
|
| 31 |
-
Curated copies of
|
| 32 |
images as stored in the OphGen image store (de-identified, de-duplicated, native resolution), the train / validation / test
|
| 33 |
split used in the project, six captions per image (short / medium / long × English / Chinese) and all original task annotations.
|
| 34 |
|
|
@@ -50,7 +62,8 @@ In `all`, an image released by several datasets appears once per dataset (same `
|
|
| 50 |
| Config | Dataset | Modalities | Train | Validation | Test | Tasks |
|
| 51 |
| :--- | :--- | :--- | ---: | ---: | ---: | ---: |
|
| 52 |
| [IDRiD](cards/IDRiD.md) | IDRiD (segmentation and grading parts) | CFP | 382 | 42 | 107 | 2 |
|
| 53 |
-
|
|
|
|
|
| 54 |
|
| 55 |
Split rules: `official` = official split; `official_val_from_train` = 10% of the official training set (by patient) used as
|
| 56 |
validation; `labeled_7_1_2` = patient-level 7:1:2 split with a fixed seed (no official split); `pretrain_90_5_5` = 90:5:5 split
|
|
|
|
| 24 |
data_files:
|
| 25 |
- split: train
|
| 26 |
path: sidecar/IDRiD/*.parquet
|
| 27 |
+
- config_name: REFUGE2
|
| 28 |
+
data_files:
|
| 29 |
+
- split: train
|
| 30 |
+
path: data/REFUGE2/train-*.parquet
|
| 31 |
+
- split: validation
|
| 32 |
+
path: data/REFUGE2/validation-*.parquet
|
| 33 |
+
- split: test
|
| 34 |
+
path: data/REFUGE2/test-*.parquet
|
| 35 |
+
- config_name: REFUGE2-sidecar
|
| 36 |
+
data_files:
|
| 37 |
+
- split: train
|
| 38 |
+
path: sidecar/REFUGE2/*.parquet
|
| 39 |
---
|
| 40 |
|
| 41 |
# OphGen — curated public ophthalmic datasets
|
| 42 |
|
| 43 |
+
Curated copies of 2 public ophthalmic imaging datasets, prepared for the OphGen ophthalmic generative-model project:
|
| 44 |
images as stored in the OphGen image store (de-identified, de-duplicated, native resolution), the train / validation / test
|
| 45 |
split used in the project, six captions per image (short / medium / long × English / Chinese) and all original task annotations.
|
| 46 |
|
|
|
|
| 62 |
| Config | Dataset | Modalities | Train | Validation | Test | Tasks |
|
| 63 |
| :--- | :--- | :--- | ---: | ---: | ---: | ---: |
|
| 64 |
| [IDRiD](cards/IDRiD.md) | IDRiD (segmentation and grading parts) | CFP | 382 | 42 | 107 | 2 |
|
| 65 |
+
| [REFUGE2](cards/REFUGE2.md) | REFUGE2 | CFP | 395 | 402 | 403 | 3 |
|
| 66 |
+
| **Total** | | | 777 | 444 | 510 | |
|
| 67 |
|
| 68 |
Split rules: `official` = official split; `official_val_from_train` = 10% of the official training set (by patient) used as
|
| 69 |
validation; `labeled_7_1_2` = patient-level 7:1:2 split with a fixed seed (no official split); `pretrain_90_5_5` = 90:5:5 split
|