Commit ·
decf4df
1
Parent(s): b7111bf
Drop redundant cbcl_aligned source (identical to cbcl by construction)
Browse files- README.md +11 -8
- data/train-00000-of-00001.parquet +2 -2
README.md
CHANGED
|
@@ -22,7 +22,7 @@ classifiers).
|
|
| 22 |
|
| 23 |
| Split | Rows | Faces / Non-faces |
|
| 24 |
|---|---|---|
|
| 25 |
-
| `train` |
|
| 26 |
| `test` | 24,045 | 472 / 23,573 (CBCL benchmark) |
|
| 27 |
| `negatives` | 29,879 | — / 29,879 (Caltech-256) |
|
| 28 |
|
|
@@ -47,18 +47,21 @@ negs = ds["negatives"]
|
|
| 47 |
|
| 48 |
## Face sources (train split)
|
| 49 |
|
| 50 |
-
|
| 51 |
-
|
| 52 |
|
| 53 |
-
- `
|
| 54 |
-
- `
|
| 55 |
positions, face fills the 48×48 frame).
|
| 56 |
|
|
|
|
|
|
|
| 57 |
| `source` | Label | Count | Notes |
|
| 58 |
|---|---|---|---|
|
| 59 |
-
| `celeba`
|
| 60 |
-
| `
|
| 61 |
-
| `cbcl`
|
|
|
|
| 62 |
|
| 63 |
Picking `source == "cbcl"` from train returns **both faces and non-faces** —
|
| 64 |
filter by `label` to pick.
|
|
|
|
| 22 |
|
| 23 |
| Split | Rows | Faces / Non-faces |
|
| 24 |
|---|---|---|
|
| 25 |
+
| `train` | 106,977 | 102,429 / 4,548 |
|
| 26 |
| `test` | 24,045 | 472 / 23,573 (CBCL benchmark) |
|
| 27 |
| `negatives` | 29,879 | — / 29,879 (Caltech-256) |
|
| 28 |
|
|
|
|
| 47 |
|
| 48 |
## Face sources (train split)
|
| 49 |
|
| 50 |
+
CelebA ships in **two paired variants at the same index** — same person,
|
| 51 |
+
two crop styles:
|
| 52 |
|
| 53 |
+
- `celeba` — loose portrait framing (hair and jaw visible).
|
| 54 |
+
- `celeba_aligned` — tight Viola-Jones crop (eyes pinned to fixed pixel
|
| 55 |
positions, face fills the 48×48 frame).
|
| 56 |
|
| 57 |
+
CBCL is already canonically aligned, so it has a single variant.
|
| 58 |
+
|
| 59 |
| `source` | Label | Count | Notes |
|
| 60 |
|---|---|---|---|
|
| 61 |
+
| `celeba` | 1 | 50,000 | Loose portrait. Filtered for frontal pose using the manual CelebA landmarks. |
|
| 62 |
+
| `celeba_aligned` | 1 | 50,000 | Same 50,000 faces tightly warped to CBCL-matched geometry. |
|
| 63 |
+
| `cbcl` | 1 | 2,429 | MIT CBCL #1, upsampled from native 19×19. Already canonically aligned. |
|
| 64 |
+
| `cbcl` | 0 | 4,548 | CBCL non-faces — matched-domain seed for stage 1 of cascade training. |
|
| 65 |
|
| 66 |
Picking `source == "cbcl"` from train returns **both faces and non-faces** —
|
| 67 |
filter by `label` to pick.
|
data/train-00000-of-00001.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:13aeb0a839ee7bb9e137ae4cfb543c52966044893574daf30dfae71064dd50b6
|
| 3 |
+
size 186679425
|