Republish as unified Parquet table (segmentation + pVQA)

README.md

@@ -1,112 +1,198 @@
 ---
 license: cc-by-4.0
-task_categories:
-  - image-segmentation
 language:
   - en
+pretty_name: Puzzle Perception (Segmentation + pVQA)
+size_categories:
+  - 10K<n<100K
+task_categories:
+  - image-segmentation
+  - visual-question-answering
 tags:
   - puzzle-perception
   - segmentation
+  - pvqa
   - chess
   - maze
   - tower-of-hanoi
+  - nqueens
   - synthetic
-pretty_name: Puzzle Perception Segmentation
-size_categories:
-  - 1K<n<10K
+configs:
+  - config_name: default
+    data_files:
+      - split: train
+        path: data/train-*.parquet
+      - split: val
+        path: data/val-*.parquet
+      - split: test
+        path: data/test-*.parquet
 ---
 
-# 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
+# Puzzle Perception — Segmentation + pVQA
 
-Masks are 8-bit PNGs with pixel values in `[0, 29]`. The mapping (also
-shipped as `classes.yaml`):
+A single table over two tasks on synthetic puzzle images:
 
-| Source task | Class ids | Count |
-|---|---|---:|
-| maze   |  0..7    |  8 |
-| chess  |  8..22   | 15 |
-| hanoi  | 23..29   |  7 |
+- **Segmentation** — per-pixel masks over chess, maze and tower-of-hanoi under one
+  unified 30-class label space.
+- **pVQA** — multiple-choice perception probes over chess and N-Queens boards.
 
-Per-class names and loss weights (cross-entropy and mIoU) are documented
-in [`classes.yaml`](./classes.yaml).
+Every row carries the same 11 columns; the `type` column says which task it
+belongs to, and columns that do not apply are `null`.
 
-## Layout
+```python
+from datasets import load_dataset
 
-This repository ships each split as a single uncompressed `.tar` archive
-(PNGs are already compressed). After extraction the on-disk layout is::
+ds = load_dataset("PuzzleBench/Puzzle_Perception", split="test")
+pvqa = ds.filter(lambda r: r["type"] == "pvqa")
+seg = ds.filter(lambda r: r["type"] == "segmentation")
 
-    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}/
+row = seg[0]
+row["image"]                 # PIL.Image, 512x512 RGB
+row["segmentation_labels"]   # e.g. ['black_square', 'white_queen', ...]
+```
 
-Repo contents:
+## Columns
+
+| Column | Type | Segmentation rows | pVQA rows |
+|---|---|---|---|
+| `id` | `int64` | unique, contiguous from 0 | unique, contiguous from 0 |
+| `image` | `Image` | 512×512 RGB | 512×512 RGB |
+| `mask` | `string` | path to a mask PNG in this repo | `null` |
+| `puzzle` | `string` | `chess` / `maze` / `hanoi` | `chess` / `nqueens` |
+| `image_path` | `string` | source-tree path (provenance) | source-tree path (provenance) |
+| `type` | `string` | `segmentation` | `pvqa` |
+| `segmentation_labels` | `list<string>` | class names present in the mask | `null` |
+| `question` | `string` | `null` | probe question text |
+| `answer` | `string` | `null` | correct option, as text |
+| `question_id` | `string` | `null` | `q1`…`q8`, joins `pvqa_questions.yaml` |
+| `options` | `list<string>` | `null` | the full answer space |
+
+`image` is embedded in the Parquet as PNG bytes, so it renders in the viewer and
+`load_dataset` hands back a `PIL.Image`. `mask` is instead a **path** to a real
+file shipped under `masks/`, which keeps label maps — whose pixel values are
+`0..29` and therefore look almost black — out of the preview:
 
-| 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 |
+```python
+from huggingface_hub import hf_hub_download
+import numpy as np
+from PIL import Image
 
-Each sample's filename encodes its source task (e.g. `chess_000625.png`,
-`maze_00012.png`, `hanoi_00042.png`).
+p = hf_hub_download("PuzzleBench/Puzzle_Perception", row["mask"], repo_type="dataset")
+mask = np.array(Image.open(p))        # (512, 512) uint8, values in [0, 29]
+```
 
-## Usage
+## Splits
 
-```python
-from huggingface_hub import snapshot_download
-from pathlib import Path
-import tarfile, numpy as np
-from PIL import Image
+| Split | Total | Segmentation | pVQA | Per-puzzle |
+|---|---:|---:|---:|---|
+| `train` | 6000 | 6000 | 0 | seg: chess 2000, hanoi 2000, maze 2000 |
+| `val` | 1500 | 1500 | 0 | seg: chess 500, hanoi 500, maze 500 |
+| `test` | 2700 | 1500 | 1200 | seg: chess 500, hanoi 500, maze 500 · pVQA: chess 800, nqueens 400 |
+| **total** | **10200** | **9000** | **1200** |  |
 
-root = Path("Puzzle_Perception_data")
-snapshot_download(
-    repo_id="PuzzleBench/Puzzle_Perception",
-    repo_type="dataset",
-    local_dir=str(root),
-)
+pVQA appears **only** in `test`, and its rows are written before the
+segmentation rows of that split.
 
-for split in ("train", "val", "test"):
-    with tarfile.open(root / f"{split}.tar") as tf:
-        tf.extractall(root)
+## Segmentation labels
 
-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
-```
+Masks are 8-bit PNGs whose pixel values **are** these class ids — no remapping
+at load time.
 
-The PuzzleBench project ships a `download.py` helper that does the
-snapshot + extract in one step, plus a `CombinedSegmentationDataset`
-PyTorch wrapper.
+| Puzzle | Class ids | Count |
+|---|---|---:|
+| maze | 0–7 | 8 |
+| chess | 8–22 | 15 |
+| hanoi | 23–29 | 7 |
+
+| Class id | Name | Puzzle |
+|---:|---|---|
+| 0 | `wall` | maze |
+| 1 | `path` | maze |
+| 2 | `start` | maze |
+| 3 | `dest_a` | maze |
+| 4 | `dest_b` | maze |
+| 5 | `dest_c` | maze |
+| 6 | `dest_d` | maze |
+| 7 | `dest_e` | maze |
+| 8 | `background` | chess |
+| 9 | `white_square` | chess |
+| 10 | `black_square` | chess |
+| 11 | `white_pawn` | chess |
+| 12 | `white_knight` | chess |
+| 13 | `white_bishop` | chess |
+| 14 | `white_rook` | chess |
+| 15 | `white_queen` | chess |
+| 16 | `white_king` | chess |
+| 17 | `black_pawn` | chess |
+| 18 | `black_knight` | chess |
+| 19 | `black_bishop` | chess |
+| 20 | `black_rook` | chess |
+| 21 | `black_queen` | chess |
+| 22 | `black_king` | chess |
+| 23 | `background` | hanoi |
+| 24 | `peg` | hanoi |
+| 25 | `disk_1` | hanoi |
+| 26 | `disk_2` | hanoi |
+| 27 | `disk_3` | hanoi |
+| 28 | `disk_4` | hanoi |
+| 29 | `disk_5` | hanoi |
+
+Note `background` appears twice — chess id 8 and hanoi id 23 are distinct
+classes in this unified namespace — so `segmentation_labels` names should be
+read together with `puzzle`. Per-class loss weights live in `classes.yaml`.
+
+## pVQA answer spaces
+
+| Puzzle | `question_id` | Question | `options` | Rows |
+|---|---|---|---|---:|
+| chess | `q1` | Which quarter of the board is the white queen in? A=top-left B=top-right C=bottom-left D=bottom-right. | `[A, B, C, D]` | 100 |
+| chess | `q2` | Which quarter of the board is the white bishop in? A=top-left B=top-right C=bottom-left D=bottom-right. | `[A, B, C, D]` | 100 |
+| chess | `q3` | Is the leftmost white pawn in the top half or the bottom half of the board? | `[top, bottom]` | 100 |
+| chess | `q4` | Is the white king above or below the black knight? | `[above, below]` | 100 |
+| chess | `q5` | Are the white rook and the black king in the same row or the same column? | `[Yes, No]` | 100 |
+| chess | `q6` | Are the white king and the black king on adjacent (touching) squares? | `[Yes, No]` | 100 |
+| chess | `q7` | Is the white queen on the same rank, file, or diagonal as the black king? | `[Yes, No]` | 100 |
+| chess | `q8` | Is there a white queen on the board? | `[Yes, No]` | 100 |
+| nqueens | `q1` | Is the leftmost queen in the top half or the bottom half of the board? | `[top, bottom]` | 100 |
+| nqueens | `q2` | Is the topmost queen in the left half or the right half of the board? | `[left, right]` | 100 |
+| nqueens | `q3` | Is the leftmost queen above or below the rightmost queen? | `[above, below]` | 100 |
+| nqueens | `q4` | Is the leftmost queen in the top, middle, or bottom third of the board? | `[top, middle, bottom]` | 100 |
+
+Answer spaces are not uniform: 2-way, 3-way, 4-way all occur, with `chess/q1`, `chess/q2` and `nqueens/q4` non-binary.
+That is why `options` is a column rather than an assumed `[Yes, No]`.
+
+Questions are deliberately coordinate-free ("the leftmost queen", not "the queen
+at row 4"), so a model must locate the referenced piece before it can answer.
+Full specs, including how each answer was derived from the board, are in
+`pvqa_questions.yaml`.
+
+## Notes and caveats
+
+- **The two tasks do not share images.** pVQA chess boards are separately
+  rendered, not the segmentation chess images, and N-Queens is not part of the
+  30-class label space. No image currently carries both a mask and a question.
+- **`image_path` repeats for N-Queens pVQA.** Each of the 100 boards is asked 4
+  questions, so 4 rows share an `image_path` with distinct `id`s. Chess pVQA asks
+  one question per board.
+- **Published N-Queens images are a derived render.** The sources are 1600×1600
+  JPEG, downscaled here to 512×512 with `PIL` `thumbnail(..., BILINEAR)` — the
+  same call the evaluation code applies, so these are the pixels models actually
+  saw. The originals remain the reference copy in the project repository.
+- Segmentation splits are sampled from the source datasets with a fixed seed
+  (`42`) and are reproducible.
+- `manifest.csv` maps `unified_id` to source task and split; `classes.yaml`
+  carries the class map and loss weights.
+
+## Files
+
+| Path | Contents |
+|---|---|
+| `data/*.parquet` | the table — 10200 rows |
+| `masks/{train,val,test}/*.png` | 9000 label maps, 8-bit, values `0..29` |
+| `classes.yaml` | 30-class map + per-class loss weights |
+| `manifest.csv` | `unified_id,source_task,source_id,split` |
+| `pvqa_questions.yaml` | published question specs, nested by task |
 
 ## License
 
-This dataset is released under **CC BY 4.0**. If you use it, please cite
-the PuzzleBench project.
+**CC BY 4.0.** If you use this dataset, please cite the PuzzleBench project.

val.tar → data/test-00000-of-00002.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:90cd8cafe864b6141ef8b4bc5b23d31b58f0ddb00f4bc7dea67559808adaf39b
-size 547921920
+oid sha256:e77c23d8d7ce22f558b3004cd2ac886d3ac662f4244972ca0290f475e1cebd75
+size 469322308

train.tar → data/test-00001-of-00002.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4d16dd80b5347827b0a38d284f9e282802776c5f728ebe326c3f1e772661de2f
-size 2218414080
+oid sha256:8ef97051323ed9047975e9cda79f4d9598ca68bd988bfd9e29c54696a32020bf
+size 140935857

data/train-00000-of-00005.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac79221476342a4525910c182c9b228ec96a6b6e099667dd3d6bff35c2f50112
+size 503330369

data/train-00001-of-00005.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c3f4d3cda889d4b1376721436b30d77f32e88a3747a8eb1c8feef93d2f0ec09
+size 503173685

data/train-00002-of-00005.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc6f8282f57344218f1daa42d6dddb4005038baa0145dd30e6137ac5be4dca2a
+size 503037550

data/train-00003-of-00005.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d13609e471db655c529292b652e1374041ff619e25770378d4642946c060a9b5
+size 503000353

data/train-00004-of-00005.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f47a8212ce759c6afa312d2de760f7cc7d8875d39cc2365c48e077ecbe46eaa
+size 179462285

data/val-00000-of-00002.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1c9b94a83d4f02a82f018a5f634fde689959e73b583cfad694b92aa029162fd
+size 503323406

data/val-00001-of-00002.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a210dd6bcb73140caf6acd78ea4be70d345fb99dea1aea60952bd1d1c10ea875
+size 38051712