Update README.md
Browse files
README.md
CHANGED
|
@@ -165,27 +165,70 @@ A refreshed, Parquet-formatted metadata dump of Danbooru, current as of January
|
|
| 165 |
|
| 166 |
## Uses
|
| 167 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 168 |
|
| 169 |
This dataset can be used to:
|
| 170 |
- Retrieve the full Danbooru image set via the metadata’s URLs
|
| 171 |
- Train or fine-tune an image tagger
|
| 172 |
- Compare against previous metadata versions to track changes, tag evolution, and historical trends
|
| 173 |
|
| 174 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 175 |
|
| 176 |
```python
|
| 177 |
-
|
|
|
|
| 178 |
|
| 179 |
-
|
| 180 |
-
|
| 181 |
-
```
|
| 182 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 183 |
|
| 184 |
## Dataset Structure
|
| 185 |
|
| 186 |
Below is a partial overview of the DataFrame columns, derived directly from the Danbooru JSONs:
|
| 187 |
|
| 188 |
-
**column values preview:**
|
| 189 |
|
| 190 |
```python
|
| 191 |
import unibox as ub
|
|
@@ -333,495 +376,6 @@ Index(['approver_id', 'bit_flags', 'created_at', 'down_score', 'fav_count',
|
|
| 333 |
</div>
|
| 334 |
|
| 335 |
|
| 336 |
-
**memory size preview**:
|
| 337 |
-
|
| 338 |
-
<div>
|
| 339 |
-
<style scoped>
|
| 340 |
-
.dataframe tbody tr th:only-of-type {
|
| 341 |
-
vertical-align: middle;
|
| 342 |
-
}
|
| 343 |
-
.dataframe tbody tr th {
|
| 344 |
-
vertical-align: top;
|
| 345 |
-
}
|
| 346 |
-
.dataframe thead th {
|
| 347 |
-
text-align: right;
|
| 348 |
-
}
|
| 349 |
-
</style>
|
| 350 |
-
<table border="1" class="dataframe">
|
| 351 |
-
<thead>
|
| 352 |
-
<tr style="text-align: right;">
|
| 353 |
-
<th></th>
|
| 354 |
-
<th>Column</th>
|
| 355 |
-
<th>Memory Usage</th>
|
| 356 |
-
<th>Readable Memory Usage</th>
|
| 357 |
-
<th>Dtype</th>
|
| 358 |
-
</tr>
|
| 359 |
-
</thead>
|
| 360 |
-
<tbody>
|
| 361 |
-
<tr>
|
| 362 |
-
<th>37</th>
|
| 363 |
-
<td>media_asset_variants</td>
|
| 364 |
-
<td>6504773310</td>
|
| 365 |
-
<td>6.06 GB</td>
|
| 366 |
-
<td>object</td>
|
| 367 |
-
</tr>
|
| 368 |
-
<tr>
|
| 369 |
-
<th>50</th>
|
| 370 |
-
<td>tag_string</td>
|
| 371 |
-
<td>4010312632</td>
|
| 372 |
-
<td>3.73 GB</td>
|
| 373 |
-
<td>object</td>
|
| 374 |
-
</tr>
|
| 375 |
-
<tr>
|
| 376 |
-
<th>54</th>
|
| 377 |
-
<td>tag_string_general</td>
|
| 378 |
-
<td>3219096004</td>
|
| 379 |
-
<td>3.00 GB</td>
|
| 380 |
-
<td>object</td>
|
| 381 |
-
</tr>
|
| 382 |
-
<tr>
|
| 383 |
-
<th>19</th>
|
| 384 |
-
<td>large_file_url</td>
|
| 385 |
-
<td>1120901064</td>
|
| 386 |
-
<td>1.04 GB</td>
|
| 387 |
-
<td>object</td>
|
| 388 |
-
</tr>
|
| 389 |
-
<tr>
|
| 390 |
-
<th>7</th>
|
| 391 |
-
<td>file_url</td>
|
| 392 |
-
<td>1091768146</td>
|
| 393 |
-
<td>1.02 GB</td>
|
| 394 |
-
<td>object</td>
|
| 395 |
-
</tr>
|
| 396 |
-
<tr>
|
| 397 |
-
<th>40</th>
|
| 398 |
-
<td>preview_file_url</td>
|
| 399 |
-
<td>1083377159</td>
|
| 400 |
-
<td>1.01 GB</td>
|
| 401 |
-
<td>object</td>
|
| 402 |
-
</tr>
|
| 403 |
-
<tr>
|
| 404 |
-
<th>43</th>
|
| 405 |
-
<td>source</td>
|
| 406 |
-
<td>1028480196</td>
|
| 407 |
-
<td>980.84 MB</td>
|
| 408 |
-
<td>object</td>
|
| 409 |
-
</tr>
|
| 410 |
-
<tr>
|
| 411 |
-
<th>34</th>
|
| 412 |
-
<td>media_asset_pixel_hash</td>
|
| 413 |
-
<td>766839397</td>
|
| 414 |
-
<td>731.32 MB</td>
|
| 415 |
-
<td>object</td>
|
| 416 |
-
</tr>
|
| 417 |
-
<tr>
|
| 418 |
-
<th>33</th>
|
| 419 |
-
<td>media_asset_md5</td>
|
| 420 |
-
<td>749456967</td>
|
| 421 |
-
<td>714.74 MB</td>
|
| 422 |
-
<td>object</td>
|
| 423 |
-
</tr>
|
| 424 |
-
<tr>
|
| 425 |
-
<th>23</th>
|
| 426 |
-
<td>md5</td>
|
| 427 |
-
<td>749456967</td>
|
| 428 |
-
<td>714.74 MB</td>
|
| 429 |
-
<td>object</td>
|
| 430 |
-
</tr>
|
| 431 |
-
<tr>
|
| 432 |
-
<th>2</th>
|
| 433 |
-
<td>created_at</td>
|
| 434 |
-
<td>740990878</td>
|
| 435 |
-
<td>706.66 MB</td>
|
| 436 |
-
<td>object</td>
|
| 437 |
-
</tr>
|
| 438 |
-
<tr>
|
| 439 |
-
<th>36</th>
|
| 440 |
-
<td>media_asset_updated_at</td>
|
| 441 |
-
<td>740990878</td>
|
| 442 |
-
<td>706.66 MB</td>
|
| 443 |
-
<td>object</td>
|
| 444 |
-
</tr>
|
| 445 |
-
<tr>
|
| 446 |
-
<th>24</th>
|
| 447 |
-
<td>media_asset_created_at</td>
|
| 448 |
-
<td>740990878</td>
|
| 449 |
-
<td>706.66 MB</td>
|
| 450 |
-
<td>object</td>
|
| 451 |
-
</tr>
|
| 452 |
-
<tr>
|
| 453 |
-
<th>57</th>
|
| 454 |
-
<td>updated_at</td>
|
| 455 |
-
<td>740990878</td>
|
| 456 |
-
<td>706.66 MB</td>
|
| 457 |
-
<td>object</td>
|
| 458 |
-
</tr>
|
| 459 |
-
<tr>
|
| 460 |
-
<th>52</th>
|
| 461 |
-
<td>tag_string_character</td>
|
| 462 |
-
<td>735974700</td>
|
| 463 |
-
<td>701.88 MB</td>
|
| 464 |
-
<td>object</td>
|
| 465 |
-
</tr>
|
| 466 |
-
<tr>
|
| 467 |
-
<th>55</th>
|
| 468 |
-
<td>tag_string_meta</td>
|
| 469 |
-
<td>734235943</td>
|
| 470 |
-
<td>700.22 MB</td>
|
| 471 |
-
<td>object</td>
|
| 472 |
-
</tr>
|
| 473 |
-
<tr>
|
| 474 |
-
<th>53</th>
|
| 475 |
-
<td>tag_string_copyright</td>
|
| 476 |
-
<td>664510146</td>
|
| 477 |
-
<td>633.73 MB</td>
|
| 478 |
-
<td>object</td>
|
| 479 |
-
</tr>
|
| 480 |
-
<tr>
|
| 481 |
-
<th>51</th>
|
| 482 |
-
<td>tag_string_artist</td>
|
| 483 |
-
<td>588454689</td>
|
| 484 |
-
<td>561.19 MB</td>
|
| 485 |
-
<td>object</td>
|
| 486 |
-
</tr>
|
| 487 |
-
<tr>
|
| 488 |
-
<th>27</th>
|
| 489 |
-
<td>media_asset_file_key</td>
|
| 490 |
-
<td>557435694</td>
|
| 491 |
-
<td>531.61 MB</td>
|
| 492 |
-
<td>object</td>
|
| 493 |
-
</tr>
|
| 494 |
-
<tr>
|
| 495 |
-
<th>35</th>
|
| 496 |
-
<td>media_asset_status</td>
|
| 497 |
-
<td>542818899</td>
|
| 498 |
-
<td>517.67 MB</td>
|
| 499 |
-
<td>object</td>
|
| 500 |
-
</tr>
|
| 501 |
-
<tr>
|
| 502 |
-
<th>26</th>
|
| 503 |
-
<td>media_asset_file_ext</td>
|
| 504 |
-
<td>516984380</td>
|
| 505 |
-
<td>493.03 MB</td>
|
| 506 |
-
<td>object</td>
|
| 507 |
-
</tr>
|
| 508 |
-
<tr>
|
| 509 |
-
<th>5</th>
|
| 510 |
-
<td>file_ext</td>
|
| 511 |
-
<td>516984380</td>
|
| 512 |
-
<td>493.03 MB</td>
|
| 513 |
-
<td>object</td>
|
| 514 |
-
</tr>
|
| 515 |
-
<tr>
|
| 516 |
-
<th>41</th>
|
| 517 |
-
<td>rating</td>
|
| 518 |
-
<td>499738034</td>
|
| 519 |
-
<td>476.59 MB</td>
|
| 520 |
-
<td>object</td>
|
| 521 |
-
</tr>
|
| 522 |
-
<tr>
|
| 523 |
-
<th>21</th>
|
| 524 |
-
<td>last_commented_at</td>
|
| 525 |
-
<td>259587166</td>
|
| 526 |
-
<td>247.56 MB</td>
|
| 527 |
-
<td>object</td>
|
| 528 |
-
</tr>
|
| 529 |
-
<tr>
|
| 530 |
-
<th>20</th>
|
| 531 |
-
<td>last_comment_bumped_at</td>
|
| 532 |
-
<td>252905426</td>
|
| 533 |
-
<td>241.19 MB</td>
|
| 534 |
-
<td>object</td>
|
| 535 |
-
</tr>
|
| 536 |
-
<tr>
|
| 537 |
-
<th>22</th>
|
| 538 |
-
<td>last_noted_at</td>
|
| 539 |
-
<td>247393688</td>
|
| 540 |
-
<td>235.93 MB</td>
|
| 541 |
-
<td>object</td>
|
| 542 |
-
</tr>
|
| 543 |
-
<tr>
|
| 544 |
-
<th>1</th>
|
| 545 |
-
<td>bit_flags</td>
|
| 546 |
-
<td>68929384</td>
|
| 547 |
-
<td>65.74 MB</td>
|
| 548 |
-
<td>int64</td>
|
| 549 |
-
</tr>
|
| 550 |
-
<tr>
|
| 551 |
-
<th>49</th>
|
| 552 |
-
<td>tag_count_meta</td>
|
| 553 |
-
<td>68929384</td>
|
| 554 |
-
<td>65.74 MB</td>
|
| 555 |
-
<td>int64</td>
|
| 556 |
-
</tr>
|
| 557 |
-
<tr>
|
| 558 |
-
<th>42</th>
|
| 559 |
-
<td>score</td>
|
| 560 |
-
<td>68929384</td>
|
| 561 |
-
<td>65.74 MB</td>
|
| 562 |
-
<td>int64</td>
|
| 563 |
-
</tr>
|
| 564 |
-
<tr>
|
| 565 |
-
<th>38</th>
|
| 566 |
-
<td>parent_id</td>
|
| 567 |
-
<td>68929384</td>
|
| 568 |
-
<td>65.74 MB</td>
|
| 569 |
-
<td>float64</td>
|
| 570 |
-
</tr>
|
| 571 |
-
<tr>
|
| 572 |
-
<th>39</th>
|
| 573 |
-
<td>pixiv_id</td>
|
| 574 |
-
<td>68929384</td>
|
| 575 |
-
<td>65.74 MB</td>
|
| 576 |
-
<td>float64</td>
|
| 577 |
-
</tr>
|
| 578 |
-
<tr>
|
| 579 |
-
<th>30</th>
|
| 580 |
-
<td>media_asset_image_height</td>
|
| 581 |
-
<td>68929384</td>
|
| 582 |
-
<td>65.74 MB</td>
|
| 583 |
-
<td>int64</td>
|
| 584 |
-
</tr>
|
| 585 |
-
<tr>
|
| 586 |
-
<th>29</th>
|
| 587 |
-
<td>media_asset_id</td>
|
| 588 |
-
<td>68929384</td>
|
| 589 |
-
<td>65.74 MB</td>
|
| 590 |
-
<td>int64</td>
|
| 591 |
-
</tr>
|
| 592 |
-
<tr>
|
| 593 |
-
<th>31</th>
|
| 594 |
-
<td>media_asset_image_width</td>
|
| 595 |
-
<td>68929384</td>
|
| 596 |
-
<td>65.74 MB</td>
|
| 597 |
-
<td>int64</td>
|
| 598 |
-
</tr>
|
| 599 |
-
<tr>
|
| 600 |
-
<th>28</th>
|
| 601 |
-
<td>media_asset_file_size</td>
|
| 602 |
-
<td>68929384</td>
|
| 603 |
-
<td>65.74 MB</td>
|
| 604 |
-
<td>int64</td>
|
| 605 |
-
</tr>
|
| 606 |
-
<tr>
|
| 607 |
-
<th>12</th>
|
| 608 |
-
<td>id</td>
|
| 609 |
-
<td>68929384</td>
|
| 610 |
-
<td>65.74 MB</td>
|
| 611 |
-
<td>int64</td>
|
| 612 |
-
</tr>
|
| 613 |
-
<tr>
|
| 614 |
-
<th>0</th>
|
| 615 |
-
<td>approver_id</td>
|
| 616 |
-
<td>68929384</td>
|
| 617 |
-
<td>65.74 MB</td>
|
| 618 |
-
<td>float64</td>
|
| 619 |
-
</tr>
|
| 620 |
-
<tr>
|
| 621 |
-
<th>4</th>
|
| 622 |
-
<td>fav_count</td>
|
| 623 |
-
<td>68929384</td>
|
| 624 |
-
<td>65.74 MB</td>
|
| 625 |
-
<td>int64</td>
|
| 626 |
-
</tr>
|
| 627 |
-
<tr>
|
| 628 |
-
<th>3</th>
|
| 629 |
-
<td>down_score</td>
|
| 630 |
-
<td>68929384</td>
|
| 631 |
-
<td>65.74 MB</td>
|
| 632 |
-
<td>int64</td>
|
| 633 |
-
</tr>
|
| 634 |
-
<tr>
|
| 635 |
-
<th>6</th>
|
| 636 |
-
<td>file_size</td>
|
| 637 |
-
<td>68929384</td>
|
| 638 |
-
<td>65.74 MB</td>
|
| 639 |
-
<td>int64</td>
|
| 640 |
-
</tr>
|
| 641 |
-
<tr>
|
| 642 |
-
<th>25</th>
|
| 643 |
-
<td>media_asset_duration</td>
|
| 644 |
-
<td>68929384</td>
|
| 645 |
-
<td>65.74 MB</td>
|
| 646 |
-
<td>float64</td>
|
| 647 |
-
</tr>
|
| 648 |
-
<tr>
|
| 649 |
-
<th>14</th>
|
| 650 |
-
<td>image_width</td>
|
| 651 |
-
<td>68929384</td>
|
| 652 |
-
<td>65.74 MB</td>
|
| 653 |
-
<td>int64</td>
|
| 654 |
-
</tr>
|
| 655 |
-
<tr>
|
| 656 |
-
<th>13</th>
|
| 657 |
-
<td>image_height</td>
|
| 658 |
-
<td>68929384</td>
|
| 659 |
-
<td>65.74 MB</td>
|
| 660 |
-
<td>int64</td>
|
| 661 |
-
</tr>
|
| 662 |
-
<tr>
|
| 663 |
-
<th>44</th>
|
| 664 |
-
<td>tag_count</td>
|
| 665 |
-
<td>68929384</td>
|
| 666 |
-
<td>65.74 MB</td>
|
| 667 |
-
<td>int64</td>
|
| 668 |
-
</tr>
|
| 669 |
-
<tr>
|
| 670 |
-
<th>48</th>
|
| 671 |
-
<td>tag_count_general</td>
|
| 672 |
-
<td>68929384</td>
|
| 673 |
-
<td>65.74 MB</td>
|
| 674 |
-
<td>int64</td>
|
| 675 |
-
</tr>
|
| 676 |
-
<tr>
|
| 677 |
-
<th>47</th>
|
| 678 |
-
<td>tag_count_copyright</td>
|
| 679 |
-
<td>68929384</td>
|
| 680 |
-
<td>65.74 MB</td>
|
| 681 |
-
<td>int64</td>
|
| 682 |
-
</tr>
|
| 683 |
-
<tr>
|
| 684 |
-
<th>46</th>
|
| 685 |
-
<td>tag_count_character</td>
|
| 686 |
-
<td>68929384</td>
|
| 687 |
-
<td>65.74 MB</td>
|
| 688 |
-
<td>int64</td>
|
| 689 |
-
</tr>
|
| 690 |
-
<tr>
|
| 691 |
-
<th>56</th>
|
| 692 |
-
<td>up_score</td>
|
| 693 |
-
<td>68929384</td>
|
| 694 |
-
<td>65.74 MB</td>
|
| 695 |
-
<td>int64</td>
|
| 696 |
-
</tr>
|
| 697 |
-
<tr>
|
| 698 |
-
<th>45</th>
|
| 699 |
-
<td>tag_count_artist</td>
|
| 700 |
-
<td>68929384</td>
|
| 701 |
-
<td>65.74 MB</td>
|
| 702 |
-
<td>int64</td>
|
| 703 |
-
</tr>
|
| 704 |
-
<tr>
|
| 705 |
-
<th>58</th>
|
| 706 |
-
<td>uploader_id</td>
|
| 707 |
-
<td>68929384</td>
|
| 708 |
-
<td>65.74 MB</td>
|
| 709 |
-
<td>int64</td>
|
| 710 |
-
</tr>
|
| 711 |
-
<tr>
|
| 712 |
-
<th>10</th>
|
| 713 |
-
<td>has_large</td>
|
| 714 |
-
<td>8616173</td>
|
| 715 |
-
<td>8.22 MB</td>
|
| 716 |
-
<td>bool</td>
|
| 717 |
-
</tr>
|
| 718 |
-
<tr>
|
| 719 |
-
<th>15</th>
|
| 720 |
-
<td>is_banned</td>
|
| 721 |
-
<td>8616173</td>
|
| 722 |
-
<td>8.22 MB</td>
|
| 723 |
-
<td>bool</td>
|
| 724 |
-
</tr>
|
| 725 |
-
<tr>
|
| 726 |
-
<th>16</th>
|
| 727 |
-
<td>is_deleted</td>
|
| 728 |
-
<td>8616173</td>
|
| 729 |
-
<td>8.22 MB</td>
|
| 730 |
-
<td>bool</td>
|
| 731 |
-
</tr>
|
| 732 |
-
<tr>
|
| 733 |
-
<th>17</th>
|
| 734 |
-
<td>is_flagged</td>
|
| 735 |
-
<td>8616173</td>
|
| 736 |
-
<td>8.22 MB</td>
|
| 737 |
-
<td>bool</td>
|
| 738 |
-
</tr>
|
| 739 |
-
<tr>
|
| 740 |
-
<th>11</th>
|
| 741 |
-
<td>has_visible_children</td>
|
| 742 |
-
<td>8616173</td>
|
| 743 |
-
<td>8.22 MB</td>
|
| 744 |
-
<td>bool</td>
|
| 745 |
-
</tr>
|
| 746 |
-
<tr>
|
| 747 |
-
<th>18</th>
|
| 748 |
-
<td>is_pending</td>
|
| 749 |
-
<td>8616173</td>
|
| 750 |
-
<td>8.22 MB</td>
|
| 751 |
-
<td>bool</td>
|
| 752 |
-
</tr>
|
| 753 |
-
<tr>
|
| 754 |
-
<th>8</th>
|
| 755 |
-
<td>has_active_children</td>
|
| 756 |
-
<td>8616173</td>
|
| 757 |
-
<td>8.22 MB</td>
|
| 758 |
-
<td>bool</td>
|
| 759 |
-
</tr>
|
| 760 |
-
<tr>
|
| 761 |
-
<th>9</th>
|
| 762 |
-
<td>has_children</td>
|
| 763 |
-
<td>8616173</td>
|
| 764 |
-
<td>8.22 MB</td>
|
| 765 |
-
<td>bool</td>
|
| 766 |
-
</tr>
|
| 767 |
-
<tr>
|
| 768 |
-
<th>32</th>
|
| 769 |
-
<td>media_asset_is_public</td>
|
| 770 |
-
<td>8616173</td>
|
| 771 |
-
<td>8.22 MB</td>
|
| 772 |
-
<td>bool</td>
|
| 773 |
-
</tr>
|
| 774 |
-
</tbody>
|
| 775 |
-
</table>
|
| 776 |
-
</div>
|
| 777 |
-
|
| 778 |
-
## Dataset Creation
|
| 779 |
-
|
| 780 |
-
We scraped all post IDs on Danbooru from 1 up to the latest. Some restricted tags (e.g. `loli`) were hidden by the site and require a gold account to access, so they are not present.
|
| 781 |
-
For a more complete (but older) metadata reference, you may wish to combine this with Danbooru2021 or similar previous scrapes.
|
| 782 |
-
|
| 783 |
-
The scraping process used a pool of roughly 400 IPs over six hours, ensuring consistent tag definitions. Below is a simplified example of the process used to convert the metadata into Parquet:
|
| 784 |
-
|
| 785 |
-
```python
|
| 786 |
-
import pandas as pd
|
| 787 |
-
from pandarallel import pandarallel
|
| 788 |
-
|
| 789 |
-
# Initialize pandarallel
|
| 790 |
-
pandarallel.initialize(nb_workers=4, progress_bar=True)
|
| 791 |
-
|
| 792 |
-
def flatten_dict(d, parent_key='', sep='_'):
|
| 793 |
-
"""
|
| 794 |
-
Flattens a nested dictionary.
|
| 795 |
-
"""
|
| 796 |
-
items = []
|
| 797 |
-
for k, v in d.items():
|
| 798 |
-
new_key = f"{parent_key}{sep}{k}" if parent_key else k
|
| 799 |
-
if isinstance(v, dict):
|
| 800 |
-
items.extend(flatten_dict(v, new_key, sep=sep).items())
|
| 801 |
-
elif isinstance(v, list):
|
| 802 |
-
items.append((new_key, ', '.join(map(str, v))))
|
| 803 |
-
else:
|
| 804 |
-
items.append((new_key, v))
|
| 805 |
-
return dict(items)
|
| 806 |
-
|
| 807 |
-
def extract_all_illust_info(json_content):
|
| 808 |
-
"""
|
| 809 |
-
Parses and flattens Danbooru JSON into a pandas Series.
|
| 810 |
-
"""
|
| 811 |
-
flattened_data = flatten_dict(json_content)
|
| 812 |
-
return pd.Series(flattened_data)
|
| 813 |
-
|
| 814 |
-
def dicts_to_dataframe_parallel(dicts):
|
| 815 |
-
"""
|
| 816 |
-
Converts a list of dicts to a flattened DataFrame using pandarallel.
|
| 817 |
-
"""
|
| 818 |
-
df = pd.DataFrame(dicts)
|
| 819 |
-
flattened_df = df.parallel_apply(lambda row: extract_all_illust_info(row.to_dict()), axis=1)
|
| 820 |
-
return flattened_df
|
| 821 |
-
```
|
| 822 |
-
|
| 823 |
-
|
| 824 |
-
|
| 825 |
### Recommendations
|
| 826 |
|
| 827 |
Users should be aware of potential biases and limitations, including the presence of adult content in some tags. More details and mitigations may be needed.
|
|
|
|
| 165 |
|
| 166 |
## Uses
|
| 167 |
|
| 168 |
+
The dataset can be loaded or filtered with the Huggingface `datasets` library:
|
| 169 |
+
|
| 170 |
+
```python
|
| 171 |
+
from datasets import Dataset, load_dataset
|
| 172 |
+
|
| 173 |
+
danbooru_dataset = load_dataset("trojblue/danbooru2025-metadata", split="train")
|
| 174 |
+
df = danbooru_dataset.to_pandas()
|
| 175 |
+
```
|
| 176 |
|
| 177 |
This dataset can be used to:
|
| 178 |
- Retrieve the full Danbooru image set via the metadata’s URLs
|
| 179 |
- Train or fine-tune an image tagger
|
| 180 |
- Compare against previous metadata versions to track changes, tag evolution, and historical trends
|
| 181 |
|
| 182 |
+
|
| 183 |
+
## Dataset Creation
|
| 184 |
+
|
| 185 |
+
We scraped all post IDs on Danbooru from 1 up to the latest. Some restricted tags (e.g. `loli`) were hidden by the site and require a gold account to access, so they are not present.
|
| 186 |
+
For a more complete (but older) metadata reference, you may wish to combine this with Danbooru2021 or similar previous scrapes.
|
| 187 |
+
|
| 188 |
+
The scraping process used a pool of roughly 400 IPs over six hours, ensuring consistent tag definitions. Below is a simplified example of the process used to convert the metadata into Parquet:
|
| 189 |
|
| 190 |
```python
|
| 191 |
+
import pandas as pd
|
| 192 |
+
from pandarallel import pandarallel
|
| 193 |
|
| 194 |
+
# Initialize pandarallel
|
| 195 |
+
pandarallel.initialize(nb_workers=4, progress_bar=True)
|
|
|
|
| 196 |
|
| 197 |
+
def flatten_dict(d, parent_key='', sep='_'):
|
| 198 |
+
"""
|
| 199 |
+
Flattens a nested dictionary.
|
| 200 |
+
"""
|
| 201 |
+
items = []
|
| 202 |
+
for k, v in d.items():
|
| 203 |
+
new_key = f"{parent_key}{sep}{k}" if parent_key else k
|
| 204 |
+
if isinstance(v, dict):
|
| 205 |
+
items.extend(flatten_dict(v, new_key, sep=sep).items())
|
| 206 |
+
elif isinstance(v, list):
|
| 207 |
+
items.append((new_key, ', '.join(map(str, v))))
|
| 208 |
+
else:
|
| 209 |
+
items.append((new_key, v))
|
| 210 |
+
return dict(items)
|
| 211 |
+
|
| 212 |
+
def extract_all_illust_info(json_content):
|
| 213 |
+
"""
|
| 214 |
+
Parses and flattens Danbooru JSON into a pandas Series.
|
| 215 |
+
"""
|
| 216 |
+
flattened_data = flatten_dict(json_content)
|
| 217 |
+
return pd.Series(flattened_data)
|
| 218 |
+
|
| 219 |
+
def dicts_to_dataframe_parallel(dicts):
|
| 220 |
+
"""
|
| 221 |
+
Converts a list of dicts to a flattened DataFrame using pandarallel.
|
| 222 |
+
"""
|
| 223 |
+
df = pd.DataFrame(dicts)
|
| 224 |
+
flattened_df = df.parallel_apply(lambda row: extract_all_illust_info(row.to_dict()), axis=1)
|
| 225 |
+
return flattened_df
|
| 226 |
+
```
|
| 227 |
|
| 228 |
## Dataset Structure
|
| 229 |
|
| 230 |
Below is a partial overview of the DataFrame columns, derived directly from the Danbooru JSONs:
|
| 231 |
|
|
|
|
| 232 |
|
| 233 |
```python
|
| 234 |
import unibox as ub
|
|
|
|
| 376 |
</div>
|
| 377 |
|
| 378 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 379 |
### Recommendations
|
| 380 |
|
| 381 |
Users should be aware of potential biases and limitations, including the presence of adult content in some tags. More details and mitigations may be needed.
|