Upload folder using huggingface_hub

README.md

@@ -0,0 +1,112 @@
+---
+license: cc-by-4.0
+task_categories:
+  - image-segmentation
+language:
+  - en
+tags:
+  - puzzle-perception
+  - segmentation
+  - chess
+  - maze
+  - tower-of-hanoi
+  - synthetic
+pretty_name: Puzzle Perception Segmentation
+size_categories:
+  - 1K<n<10K
+---
+
+# Puzzle Perception Segmentation
+
+A combined per-pixel segmentation dataset over three puzzle domains —
+chess, maze, and tower-of-hanoi — under a single unified 30-class label
+space. The published split is a deterministically sampled subset of three
+larger raw source datasets, sized to match a standard
+segmentation-probe training recipe.
+
+This dataset is used by the **PuzzleBench** project. The loader and full
+project code live at the project's repository; this HuggingFace repo
+ships only the bulk image + mask payload plus the class definitions.
+
+## Splits
+
+| Split | Per source task | Total |
+|---|---:|---:|
+| `train` | 2000 | 6000 |
+| `val`   |  500 | 1500 |
+| `test`  |  500 | 1500 |
+
+Splits are produced from the source datasets with a fixed seed (`42`),
+so the subset is reproducible.
+
+## Class space
+
+Masks are 8-bit PNGs with pixel values in `[0, 29]`. The mapping (also
+shipped as `classes.yaml`):
+
+| Source task | Class ids | Count |
+|---|---|---:|
+| maze   |  0..7    |  8 |
+| chess  |  8..22   | 15 |
+| hanoi  | 23..29   |  7 |
+
+Per-class names and loss weights (cross-entropy and mIoU) are documented
+in [`classes.yaml`](./classes.yaml).
+
+## Layout
+
+This repository ships each split as a single uncompressed `.tar` archive
+(PNGs are already compressed). After extraction the on-disk layout is::
+
+    data/
+    ├── classes.yaml
+    ├── manifest.csv                       # unified_id, source_task, source_id, split
+    ├── train/
+    │   ├── images/<task>_<id>.png         # source image (512×512 RGB)
+    │   └── masks/<task>_<id>.png          # mask  (512×512, uint8, values 0..29)
+    ├── val/{images,masks}/
+    └── test/{images,masks}/
+
+Repo contents:
+
+| File | Notes |
+|---|---|
+| `README.md`, `classes.yaml`, `manifest.csv` | Documentation + class map + per-sample index |
+| `train.tar` | 6000 (image, mask) pairs, ~2.1 GB |
+| `val.tar`   | 1500 (image, mask) pairs, ~520 MB |
+| `test.tar`  | 1500 (image, mask) pairs, ~510 MB |
+
+Each sample's filename encodes its source task (e.g. `chess_000625.png`,
+`maze_00012.png`, `hanoi_00042.png`).
+
+## Usage
+
+```python
+from huggingface_hub import snapshot_download
+from pathlib import Path
+import tarfile, numpy as np
+from PIL import Image
+
+root = Path("Puzzle_Perception_data")
+snapshot_download(
+    repo_id="PuzzleBench/Puzzle_Perception",
+    repo_type="dataset",
+    local_dir=str(root),
+)
+
+for split in ("train", "val", "test"):
+    with tarfile.open(root / f"{split}.tar") as tf:
+        tf.extractall(root)
+
+img  = np.array(Image.open(root / "train/images/chess_000625.png"))   # (512, 512, 3) uint8
+mask = np.array(Image.open(root / "train/masks/chess_000625.png"))    # (512, 512) uint8, in [8..22] for chess
+```
+
+The PuzzleBench project ships a `download.py` helper that does the
+snapshot + extract in one step, plus a `CombinedSegmentationDataset`
+PyTorch wrapper.
+
+## License
+
+This dataset is released under **CC BY 4.0**. If you use it, please cite
+the PuzzleBench project.

classes.yaml

@@ -0,0 +1,57 @@
+# Unified 30-class label space for the combined puzzle-perception segmentation
+# dataset. Class ids 0..29 are the values found in every mask PNG under data/.
+#
+# Source-task offsets:
+#     maze   →  0   ( 8 classes,  ids  0.. 7)
+#     chess  →  8   (15 classes,  ids  8..22)
+#     hanoi  → 23   ( 7 classes,  ids 23..29)
+#
+# `weight_ce`   — per-class cross-entropy weight (handles class imbalance).
+# `weight_miou` — per-class weighting for mIoU evaluation.
+# Both are inherited from the per-task definitions in the source datasets.
+#
+# `background` appears twice (chess id 8 and hanoi id 23). The two are distinct
+# classes in this unified namespace because they originate from different
+# source tasks.
+
+classes:
+  # ---- maze (offset 0) ----
+  - {id:  0, name: wall,         source_task: maze,  weight_ce: 12.0, weight_miou: 6.0}
+  - {id:  1, name: path,         source_task: maze,  weight_ce:  0.2, weight_miou: 0.3}
+  - {id:  2, name: start,        source_task: maze,  weight_ce:  6.0, weight_miou: 1.0}
+  - {id:  3, name: dest_a,       source_task: maze,  weight_ce:  6.0, weight_miou: 1.0}
+  - {id:  4, name: dest_b,       source_task: maze,  weight_ce:  6.0, weight_miou: 1.0}
+  - {id:  5, name: dest_c,       source_task: maze,  weight_ce:  6.0, weight_miou: 1.0}
+  - {id:  6, name: dest_d,       source_task: maze,  weight_ce:  6.0, weight_miou: 1.0}
+  - {id:  7, name: dest_e,       source_task: maze,  weight_ce:  6.0, weight_miou: 1.0}
+
+  # ---- chess (offset 8) ----
+  - {id:  8, name: background,   source_task: chess, weight_ce:  0.5, weight_miou: 0.5}
+  - {id:  9, name: white_square, source_task: chess, weight_ce:  0.3, weight_miou: 0.5}
+  - {id: 10, name: black_square, source_task: chess, weight_ce:  0.3, weight_miou: 0.5}
+  - {id: 11, name: white_pawn,   source_task: chess, weight_ce:  2.0, weight_miou: 1.0}
+  - {id: 12, name: white_knight, source_task: chess, weight_ce:  4.0, weight_miou: 1.0}
+  - {id: 13, name: white_bishop, source_task: chess, weight_ce:  4.0, weight_miou: 1.0}
+  - {id: 14, name: white_rook,   source_task: chess, weight_ce:  4.0, weight_miou: 1.0}
+  - {id: 15, name: white_queen,  source_task: chess, weight_ce:  6.0, weight_miou: 1.0}
+  - {id: 16, name: white_king,   source_task: chess, weight_ce:  6.0, weight_miou: 1.0}
+  - {id: 17, name: black_pawn,   source_task: chess, weight_ce:  2.0, weight_miou: 1.0}
+  - {id: 18, name: black_knight, source_task: chess, weight_ce:  4.0, weight_miou: 1.0}
+  - {id: 19, name: black_bishop, source_task: chess, weight_ce:  4.0, weight_miou: 1.0}
+  - {id: 20, name: black_rook,   source_task: chess, weight_ce:  4.0, weight_miou: 1.0}
+  - {id: 21, name: black_queen,  source_task: chess, weight_ce:  6.0, weight_miou: 1.0}
+  - {id: 22, name: black_king,   source_task: chess, weight_ce:  6.0, weight_miou: 1.0}
+
+  # ---- hanoi (offset 23) ----
+  - {id: 23, name: background,   source_task: hanoi, weight_ce:  0.5, weight_miou: 0.5}
+  - {id: 24, name: peg,          source_task: hanoi, weight_ce:  4.0, weight_miou: 1.0}
+  - {id: 25, name: disk_1,       source_task: hanoi, weight_ce:  3.0, weight_miou: 1.0}
+  - {id: 26, name: disk_2,       source_task: hanoi, weight_ce:  3.0, weight_miou: 1.0}
+  - {id: 27, name: disk_3,       source_task: hanoi, weight_ce:  3.0, weight_miou: 1.0}
+  - {id: 28, name: disk_4,       source_task: hanoi, weight_ce:  3.0, weight_miou: 1.0}
+  - {id: 29, name: disk_5,       source_task: hanoi, weight_ce:  3.0, weight_miou: 1.0}
+
+label_offsets:
+  maze:  0
+  chess: 8
+  hanoi: 23

test.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:107f02cff3d43838d53f4289f8598b56211659f171586e8eb1ffdb3a69d19786
+size 531671040

train.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d16dd80b5347827b0a38d284f9e282802776c5f728ebe326c3f1e772661de2f
+size 2218414080

val.tar

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90cd8cafe864b6141ef8b4bc5b23d31b58f0ddb00f4bc7dea67559808adaf39b
+size 547921920