Risposta commited on
Commit
c99a4af
·
verified ·
1 Parent(s): fe95c26

Update dataset card

Browse files
Files changed (1) hide show
  1. README.md +135 -0
README.md ADDED
@@ -0,0 +1,135 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: other
3
+ license_name: research-only
4
+ license_link: https://github.com/jurky123/MC_Gen2.0
5
+ pretty_name: "MC-Gen 2.0 — Minecraft / pixel-art 32x32 texture datasets"
6
+ task_categories:
7
+ - text-to-image
8
+ tags:
9
+ - minecraft
10
+ - pixel-art
11
+ - texture
12
+ - diffusion
13
+ - flow-matching
14
+ - mmap
15
+ - research
16
+ size_categories:
17
+ - 1M<n<10M
18
+ ---
19
+
20
+ # MC-Gen 2.0 — training datasets
21
+
22
+ Training-ready datasets for the **MC-Gen 2.0** project: a small pixel-space
23
+ Rectified-Flow Transformer that generates Minecraft / voxel-game style
24
+ low-resolution textures (32x32, RGBA).
25
+
26
+ The source code lives on GitHub: **https://github.com/jurky123/MC_Gen2.0**
27
+
28
+ This repository only contains the **preprocessed, training-ready** datasets.
29
+ Raw downloads (itch.io, OpenGameArt, Kaggle, Modrinth, Hugging Face mirrors)
30
+ are intentionally not included.
31
+
32
+ > **License: research / study only.** The datasets aggregate third-party game
33
+ > assets with heterogeneous licenses. They are published here for academic and
34
+ > personal study only. Do not use them commercially, and keep the provenance
35
+ > fields in the metadata for attribution. See *Provenance* below.
36
+
37
+ ## Contents
38
+
39
+ | Path | Rows | Shape / dtype | Split (train/val/test) | Purpose |
40
+ |---|---|---|---|---|
41
+ | `data/build/stage1_32_rgba` | 3,185,719 | `uint8[32,32,4]` RGBA | 3,004,255 / 91,566 / 89,898 | **Stage 1** pixel + all-MC pretraining |
42
+ | `data/build/mc_text2image32` | 1,034,057 | `uint8[32,32,3]` RGB | 941,104 / 43,184 / 49,769 | Minecraft domain (text+image), Stage 1/2 |
43
+ | `data/processed/minecraft_16x_finetune32` | 1,498 | `32x32` tiles | manifest only | **Stage 3** curated fine-tune seeds |
44
+ | `data/processed/modrinth32` | 10,818 | `32x32` tiles | manifest only | Modrinth MC textures (deduped) |
45
+
46
+ ### `stage1_32_rgba` source mix
47
+
48
+ | source_set | rows |
49
+ |---|---:|
50
+ | itch.io free pixel art | 1,859,029 |
51
+ | NathMen12 MC text-to-image | 1,034,057 |
52
+ | alucard pixel sprites | 282,044 |
53
+ | OpenGameArt (OGA-BY 3.0) | 4,555 |
54
+ | Kenney pixel assets | 4,312 |
55
+ | Kaggle pixel art | 1,722 |
56
+
57
+ Minecraft rows are fully opaque; pixel-art rows keep their original alpha
58
+ channel (transparent backgrounds are **not** composited onto black).
59
+
60
+ ## File format
61
+
62
+ Each mmap dataset directory contains:
63
+
64
+ ```text
65
+ <name>/
66
+ ├── images.uint8.mmap # raw little-endian uint8, C-order (N, H, W, C)
67
+ ├── metadata.parquet # one row per image + provenance/prompt fields
68
+ ├── splits.json # {"train": [idx...], "val": [...], "test": [...]}
69
+ └── build_summary.json # counts and source breakdown
70
+ ```
71
+
72
+ `images.uint8.mmap` is a headerless NumPy array. Load it with:
73
+
74
+ ```python
75
+ import numpy as np
76
+ n, size, channels = 3_185_719, 32, 4
77
+ imgs = np.memmap(
78
+ "data/build/stage1_32_rgba/images.uint8.mmap",
79
+ dtype=np.uint8, mode="r", shape=(n, size, size, channels),
80
+ )
81
+ ```
82
+
83
+ `metadata.parquet` columns for the unified datasets:
84
+
85
+ | column | meaning |
86
+ |---|---|
87
+ | `index` | row index into the mmap |
88
+ | `source_set` | originating processed set (see table above) |
89
+ | `weak_prompt` | short text caption (MC filename/label or generic pixel-art) |
90
+ | `source_metadata` | JSON blob with original filename, project, license, etc. |
91
+
92
+ The processed `*32` tile folders instead contain `tiles/<sha>.png` plus a
93
+ `manifest.jsonl` and `summary.json`.
94
+
95
+ ## Download
96
+
97
+ ```bash
98
+ pip install -U "huggingface_hub[hf_xet]"
99
+
100
+ # only the pretraining set
101
+ hf download Risposta/MC_Gen \
102
+ --repo-type dataset \
103
+ --include "data/build/stage1_32_rgba/*" \
104
+ --local-dir .
105
+
106
+ # everything (all four datasets)
107
+ hf download Risposta/MC_Gen --repo-type dataset --local-dir .
108
+ ```
109
+
110
+ With `--local-dir .` run from the project root, the files land exactly at the
111
+ paths the training configs expect. A convenience wrapper is provided in the
112
+ project repo as `scripts/download_dataset_hf.py`.
113
+
114
+ ## Provenance
115
+
116
+ The datasets derive from the following sources (see `source_metadata` for
117
+ per-sample details): NathMen12/16xModdedMinecraft-TextToImage, itch.io free
118
+ pixel-art asset packs, Kenney CC0 packs, evilsocket/alucard-sprites, Kaggle
119
+ ebrahimelgazar/pixel-art, nyuuzyou/OpenGameArt-OGA-BY-3.0,
120
+ James-A/Minecraft-16x-Dataset, and Modrinth resource packs.
121
+
122
+ This repository is **not affiliated with, endorsed by, or associated with**
123
+ Microsoft, Mojang, or any asset author. All trademarks belong to their
124
+ respective owners.
125
+
126
+ ## Citation
127
+
128
+ ```bibtex
129
+ @misc{mcgen2,
130
+ title = {MC-Gen 2.0: pixel-space flow transformer for Minecraft textures},
131
+ author = {Risposta},
132
+ year = {2026},
133
+ url = {https://github.com/jurky123/MC_Gen2.0}
134
+ }
135
+ ```