Kaphathy commited on
Commit
5c7df2c
·
verified ·
1 Parent(s): e6c72c3

README: 1 datasets

Browse files
Files changed (1) hide show
  1. README.md +77 -0
README.md ADDED
@@ -0,0 +1,77 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: other
3
+ license_name: private
4
+ pretty_name: OphGen curated institutional ophthalmic datasets
5
+ configs:
6
+ - config_name: all
7
+ default: true
8
+ data_files:
9
+ - split: train
10
+ path: data/*/train-*.parquet
11
+ - split: validation
12
+ path: data/*/validation-*.parquet
13
+ - split: test
14
+ path: data/*/test-*.parquet
15
+ - config_name: chengdu_ffa
16
+ data_files:
17
+ - split: train
18
+ path: data/chengdu_ffa/train-*.parquet
19
+ - split: validation
20
+ path: data/chengdu_ffa/validation-*.parquet
21
+ - split: test
22
+ path: data/chengdu_ffa/test-*.parquet
23
+ ---
24
+
25
+ # OphGen — curated institutional ophthalmic datasets
26
+
27
+ Curated copies of 1 institutional ophthalmic imaging datasets, prepared for the OphGen ophthalmic generative-model project:
28
+ images as stored in the OphGen image store (de-identified, de-duplicated, native resolution), the train / validation / test
29
+ split used in the project, six captions per image (short / medium / long × English / Chinese) and all original task annotations.
30
+
31
+ **Private hospital data: access is restricted to the OphGen project; do not redistribute or re-upload.** Images were de-identified (burned-in personal text removed); `source_paths` are paths in the hospital archives.
32
+
33
+ ## Loading
34
+
35
+ ```python
36
+ from datasets import load_dataset
37
+ ds = load_dataset("Kaphathy/OphGen-Private", "chengdu_ffa") # one dataset, its own splits and labels
38
+ side = load_dataset("Kaphathy/OphGen-Private", "chengdu_ffa-sidecar") # its other annotation files (path, size, bytes), if any
39
+ pre = load_dataset("Kaphathy/OphGen-Private", "all") # every dataset (default config), e.g. for pretraining
40
+ ```
41
+
42
+ In `all`, an image released by several datasets appears once per dataset (same `image_id`, same split); deduplicate on `image_id` if needed.
43
+
44
+ ## Datasets
45
+
46
+ | Config | Dataset | Modalities | Train | Validation | Test | Tasks |
47
+ | :--- | :--- | :--- | ---: | ---: | ---: | ---: |
48
+ | [chengdu_ffa](cards/chengdu_ffa.md) | chengdu_ffa | FFA | 3,519 | 473 | 934 | 1 |
49
+ | **Total** | | | 3,519 | 473 | 934 | |
50
+
51
+ Split rules: `official` = official split; `official_val_from_train` = 10% of the official training set (by patient) used as
52
+ 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
53
+ for images without task labels. An image with different splits across tasks takes the strictest (test > validation > train).
54
+
55
+ ## Columns (dataset configs and `all`)
56
+
57
+ | Column | Description |
58
+ | :--- | :--- |
59
+ | `image_id` | SHA-256 of the original file (stable identifier) |
60
+ | `image` | image bytes in the stored format; native resolution (no resizing) |
61
+ | `split`, `split_rule` | partition used in the project and how it was assigned |
62
+ | `modality`, `modality_name` | one of 13 standard ophthalmic modalities |
63
+ | `width`, `height`, `format`, `field_of_view`, `device`, `eye`, `sex`, `age` | metadata (empty when not documented) |
64
+ | `representation`, `octa_layer`, `scan_area_mm` | OCTA-specific fields |
65
+ | `short_en` … `long_zh` | captions built only from recorded facts; images in official test sets carry no task labels in captions (`benchmark_holdout`) |
66
+ | `source_paths` | original path(s) of the image in the hospital archive |
67
+ | `labels` | task labels of this dataset: `task`, `name` (label name, value encoding included), `value`, `num`, `official_split`, `annotation` / `label_file` (path of the annotation, embedded in `masks` or stored in the `-sidecar` config) |
68
+ | `masks` | raster annotation masks for this image (`name` matches `labels.name`, `path` matches `labels.annotation`) |
69
+
70
+ `<dataset>-sidecar` configs: `path`, `size`, `bytes` of every other annotation file of the release (tables, JSON, NPZ, MAT, XML …);
71
+ write `bytes` to `path` to restore the original layout.
72
+
73
+ ## Processing
74
+
75
+ Exact and pixel-identical duplicates were merged (labels kept); images with burned-in personal text or non-removable overlays,
76
+ undecodable or blank files were removed; images were not resized. Where the release did not document a mask value mapping, the
77
+ mapping was inferred and is stated in the label name. Full methods: OphGen supplementary methods.