Datasets:

finedet
/

owod

width int64 127 500	height int64 71 500	objects dict
500	375	{ "bbox": [ [ 262, 210, 62, 129 ], [ 164, 263, 89, 109 ], [ 4, 243, 63, 131 ], [ 240, 193, 55, 106 ], [ 276, 185, 36, 35 ] ], "category": [ 8, 8, ...
500	333	{ "bbox": [ [ 140, 49, 360, 281 ] ], "category": [ 6 ] }
500	375	{ "bbox": [ [ 68, 171, 202, 159 ], [ 149, 140, 80, 144 ], [ 284, 200, 43, 131 ], [ 257, 197, 40, 132 ] ], "category": [ 12, 14, 14, 14 ] }
500	333	{ "bbox": [ [ 155, 96, 196, 174 ] ], "category": [ 6 ] }
334	500	{ "bbox": [ [ 91, 71, 214, 402 ] ], "category": [ 1 ] }
480	364	{ "bbox": [ [ 184, 61, 95, 138 ], [ 89, 77, 314, 259 ] ], "category": [ 14, 12 ] }
500	375	{ "bbox": [ [ 230, 87, 253, 169 ], [ 10, 112, 256, 147 ] ], "category": [ 7, 7 ] }
375	500	{ "bbox": [ [ 32, 147, 339, 269 ] ], "category": [ 6 ] }
336	500	{ "bbox": [ [ 0, 234, 182, 154 ], [ 209, 35, 127, 447 ], [ 45, 81, 125, 284 ], [ 10, 180, 132, 239 ] ], "category": [ 11, 14, 14, 14 ] }
334	500	{ "bbox": [ [ 8, 229, 237, 271 ], [ 229, 219, 105, 281 ], [ 1, 177, 89, 323 ], [ 1, 0, 116, 369 ], [ 2, 1, 241, 461 ], [ 224, 0, 110, ...
500	335	{ "bbox": [ [ 195, 164, 294, 83 ] ], "category": [ 18 ] }
500	333	{ "bbox": [ [ 89, 124, 248, 88 ] ], "category": [ 6 ] }
500	375	{ "bbox": [ [ 35, 204, 145, 85 ], [ 50, 159, 100, 133 ], [ 294, 137, 156, 153 ] ], "category": [ 1, 14, 14 ] }
500	281	{ "bbox": [ [ 103, 77, 272, 106 ], [ 132, 87, 65, 36 ], [ 194, 179, 19, 50 ], [ 25, 188, 19, 50 ] ], "category": [ 0, 0, 14, 14 ] }
500	366	{ "bbox": [ [ 8, 106, 491, 157 ], [ 420, 199, 62, 27 ], [ 324, 187, 87, 36 ] ], "category": [ 0, 0, 0 ] }
360	500	{ "bbox": [ [ 115, 166, 245, 234 ], [ 140, 152, 193, 77 ] ], "category": [ 18, 18 ] }
500	375	{ "bbox": [ [ 0, 95, 191, 266 ], [ 217, 97, 248, 221 ], [ 467, 194, 33, 123 ], [ 2, 303, 498, 72 ] ], "category": [ 14, 14, 14, 10 ] }
332	500	{ "bbox": [ [ 26, 78, 293, 266 ] ], "category": [ 11 ] }
500	375	{ "bbox": [ [ 155, 88, 189, 191 ] ], "category": [ 19 ] }
500	333	{ "bbox": [ [ 362, 46, 70, 61 ], [ 215, 91, 92, 211 ], [ 163, 147, 64, 97 ] ], "category": [ 19, 14, 14 ] }
500	335	{ "bbox": [ [ 262, 31, 238, 264 ], [ 0, 35, 235, 264 ] ], "category": [ 18, 18 ] }
500	333	{ "bbox": [ [ 0, 0, 370, 330 ], [ 98, 100, 214, 113 ] ], "category": [ 8, 7 ] }
332	500	{ "bbox": [ [ 210, 233, 107, 133 ] ], "category": [ 2 ] }
500	333	{ "bbox": [ [ 14, 0, 445, 312 ], [ 438, 169, 49, 61 ] ], "category": [ 6, 8 ] }
375	500	{ "bbox": [ [ 57, 106, 234, 359 ], [ 1, 0, 301, 500 ] ], "category": [ 2, 14 ] }
500	375	{ "bbox": [ [ 404, 152, 89, 103 ], [ 274, 128, 38, 42 ], [ 359, 191, 22, 74 ], [ 398, 180, 24, 55 ], [ 269, 179, 22, 68 ], [ 293, 175, 1...
500	375	{ "bbox": [ [ 351, 137, 149, 238 ], [ 104, 0, 323, 245 ], [ 414, 60, 51, 135 ] ], "category": [ 13, 13, 14 ] }
500	333	{ "bbox": [ [ 413, 231, 45, 32 ], [ 290, 236, 43, 34 ], [ 182, 237, 45, 36 ], [ 51, 243, 48, 31 ], [ 61, 95, 46, 37 ], [ 183, 93, 42, ...
500	334	{ "bbox": [ [ 0, 136, 427, 197 ], [ 0, 61, 479, 174 ], [ 0, 26, 422, 115 ], [ 43, 6, 415, 128 ], [ 198, 5, 277, 93 ] ], "category": [ 6, 6, 6,...
500	333	{ "bbox": [ [ 273, 10, 164, 269 ], [ 183, 213, 98, 39 ] ], "category": [ 3, 3 ] }
500	375	{ "bbox": [ [ 122, 114, 257, 161 ], [ 74, 0, 354, 375 ] ], "category": [ 11, 8 ] }
496	500	{ "bbox": [ [ 0, 22, 451, 478 ] ], "category": [ 2 ] }
500	335	{ "bbox": [ [ 1, 103, 397, 231 ] ], "category": [ 11 ] }
500	375	{ "bbox": [ [ 208, 186, 20, 44 ], [ 241, 181, 33, 78 ], [ 268, 187, 27, 72 ] ], "category": [ 14, 14, 14 ] }
333	500	{ "bbox": [ [ 39, 70, 294, 403 ], [ 32, 41, 299, 95 ] ], "category": [ 13, 13 ] }
375	500	{ "bbox": [ [ 120, 142, 255, 318 ], [ 1, 153, 63, 306 ], [ 269, 154, 106, 177 ], [ 21, 142, 125, 358 ] ], "category": [ 15, 15, 3, 14 ] }
500	333	{ "bbox": [ [ 239, 0, 261, 333 ], [ 1, 0, 305, 333 ] ], "category": [ 7, 7 ] }
500	419	{ "bbox": [ [ 14, 74, 461, 338 ], [ 93, 40, 344, 198 ] ], "category": [ 11, 11 ] }
500	333	{ "bbox": [ [ 234, 157, 228, 125 ], [ 7, 165, 213, 157 ], [ 197, 155, 144, 76 ], [ 348, 144, 103, 122 ], [ 60, 161, 110, 172 ], [ 28, 165, ...
500	333	{ "bbox": [ [ 108, 32, 339, 277 ], [ 379, 1, 56, 134 ], [ 137, 13, 56, 102 ], [ 0, 19, 75, 185 ], [ 59, 47, 55, 109 ], [ 31, 50, 52, ...
500	374	{ "bbox": [ [ 19, 6, 164, 349 ], [ 97, 214, 332, 160 ], [ 331, 139, 124, 227 ], [ 21, 50, 296, 241 ] ], "category": [ 14, 14, 14, 8 ] }
500	375	{ "bbox": [ [ 124, 240, 71, 44 ], [ 80, 247, 25, 21 ], [ 90, 249, 36, 25 ] ], "category": [ 6, 6, 6 ] }
500	339	{ "bbox": [ [ 174, 124, 170, 130 ], [ 52, 2, 448, 336 ] ], "category": [ 11, 8 ] }
500	323	{ "bbox": [ [ 19, 51, 420, 234 ], [ 0, 124, 114, 110 ] ], "category": [ 18, 18 ] }
500	375	{ "bbox": [ [ 3, 76, 317, 176 ] ], "category": [ 7 ] }
500	333	{ "bbox": [ [ 184, 57, 164, 129 ], [ 39, 63, 361, 270 ] ], "category": [ 8, 14 ] }
500	375	{ "bbox": [ [ 415, 174, 85, 140 ], [ 120, 133, 232, 141 ] ], "category": [ 17, 17 ] }
500	333	{ "bbox": [ [ 402, 109, 98, 156 ], [ 290, 74, 152, 259 ], [ 1, 6, 282, 327 ] ], "category": [ 19, 14, 14 ] }
500	375	{ "bbox": [ [ 470, 173, 30, 54 ], [ 295, 148, 39, 30 ], [ 241, 145, 18, 24 ] ], "category": [ 16, 9, 16 ] }
500	254	{ "bbox": [ [ 16, 11, 466, 224 ] ], "category": [ 2 ] }
500	375	{ "bbox": [ [ 78, 232, 32, 82 ], [ 53, 230, 33, 114 ], [ 30, 232, 20, 71 ] ], "category": [ 14, 14, 14 ] }
333	500	{ "bbox": [ [ 69, 173, 208, 155 ], [ 286, 161, 30, 37 ] ], "category": [ 11, 8 ] }
500	332	{ "bbox": [ [ 276, 100, 78, 160 ] ], "category": [ 14 ] }
500	333	{ "bbox": [ [ 0, 41, 498, 243 ] ], "category": [ 0 ] }
500	375	{ "bbox": [ [ 163, 96, 118, 279 ] ], "category": [ 7 ] }
500	400	{ "bbox": [ [ 65, 120, 375, 156 ] ], "category": [ 11 ] }
500	375	{ "bbox": [ [ 250, 27, 225, 240 ], [ 21, 27, 230, 246 ] ], "category": [ 19, 19 ] }
475	480	{ "bbox": [ [ 35, 19, 395, 455 ] ], "category": [ 7 ] }
500	375	{ "bbox": [ [ 0, 25, 358, 315 ] ], "category": [ 18 ] }
500	375	{ "bbox": [ [ 424, 217, 76, 58 ], [ 435, 187, 32, 36 ], [ 381, 208, 34, 64 ], [ 352, 212, 31, 64 ] ], "category": [ 1, 14, 14, 14 ] }
334	500	{ "bbox": [ [ 69, 201, 186, 299 ], [ 250, 241, 84, 259 ], [ 0, 143, 67, 293 ], [ 0, 0, 66, 363 ], [ 73, 0, 199, 462 ], [ 251, 18, 83, ...
500	375	{ "bbox": [ [ 3, 0, 497, 249 ] ], "category": [ 3 ] }
500	333	{ "bbox": [ [ 0, 98, 475, 168 ], [ 123, 124, 49, 29 ], [ 183, 120, 43, 34 ] ], "category": [ 6, 14, 14 ] }
500	421	{ "bbox": [ [ 42, 65, 237, 298 ] ], "category": [ 5 ] }
447	500	{ "bbox": [ [ 72, 17, 334, 483 ], [ 134, 226, 204, 274 ], [ 21, 0, 54, 101 ], [ 41, 57, 96, 47 ] ], "category": [ 14, 12, 14, 12 ] }
500	371	{ "bbox": [ [ 0, 114, 487, 216 ] ], "category": [ 6 ] }
500	332	{ "bbox": [ [ 5, 55, 149, 230 ], [ 144, 57, 139, 201 ], [ 223, 53, 217, 279 ], [ 362, 95, 138, 237 ] ], "category": [ 14, 14, 14, 14 ] }
500	401	{ "bbox": [ [ 106, 145, 280, 155 ], [ 1, 112, 499, 289 ] ], "category": [ 11, 17 ] }
500	423	{ "bbox": [ [ 24, 52, 394, 347 ] ], "category": [ 2 ] }
500	375	{ "bbox": [ [ 25, 80, 440, 186 ] ], "category": [ 6 ] }
500	333	{ "bbox": [ [ 169, 45, 331, 288 ] ], "category": [ 7 ] }
374	500	{ "bbox": [ [ 154, 207, 98, 201 ] ], "category": [ 14 ] }
500	333	{ "bbox": [ [ 32, 104, 33, 42 ], [ 76, 113, 38, 36 ], [ 145, 124, 30, 26 ], [ 226, 131, 21, 23 ], [ 301, 126, 26, 27 ], [ 372, 120, 30, ...
500	333	{ "bbox": [ [ 185, 77, 134, 187 ], [ 352, 53, 27, 55 ], [ 414, 50, 22, 59 ], [ 475, 53, 17, 45 ], [ 488, 50, 12, 50 ] ], "category": [ 12, 14, ...
500	375	{ "bbox": [ [ 236, 146, 122, 45 ], [ 137, 55, 29, 15 ], [ 46, 113, 72, 12 ], [ 2, 108, 62, 14 ] ], "category": [ 6, 6, 6, 6 ] }
500	333	{ "bbox": [ [ 58, 75, 309, 191 ], [ 440, 192, 18, 21 ] ], "category": [ 3, 3 ] }
500	375	{ "bbox": [ [ 40, 72, 341, 280 ], [ 431, 102, 65, 42 ], [ 403, 113, 32, 30 ] ], "category": [ 17, 6, 6 ] }
500	375	{ "bbox": [ [ 195, 82, 72, 125 ] ], "category": [ 2 ] }
500	333	{ "bbox": [ [ 233, 47, 53, 77 ], [ 0, 15, 498, 318 ] ], "category": [ 14, 6 ] }
500	390	{ "bbox": [ [ 103, 33, 343, 357 ], [ 67, 194, 54, 94 ] ], "category": [ 6, 6 ] }
500	375	{ "bbox": [ [ 305, 226, 75, 73 ], [ 195, 142, 114, 227 ] ], "category": [ 19, 14 ] }
500	432	{ "bbox": [ [ 51, 21, 257, 307 ], [ 25, 107, 431, 289 ] ], "category": [ 14, 13 ] }
500	375	{ "bbox": [ [ 113, 153, 256, 195 ], [ 291, 48, 155, 322 ] ], "category": [ 13, 14 ] }
500	386	{ "bbox": [ [ 88, 53, 412, 333 ] ], "category": [ 14 ] }
500	333	{ "bbox": [ [ 22, 27, 458, 257 ], [ 101, 2, 60, 48 ], [ 438, 22, 62, 156 ], [ 418, 38, 26, 43 ], [ 453, 35, 34, 36 ], [ 11, 45, 93, ...
500	375	{ "bbox": [ [ 86, 99, 23, 66 ], [ 40, 113, 33, 68 ], [ 323, 147, 29, 59 ], [ 425, 156, 18, 39 ], [ 2, 90, 41, 116 ], [ 3, 27, 458, ...
375	500	{ "bbox": [ [ 0, 289, 128, 118 ], [ 93, 20, 282, 471 ] ], "category": [ 11, 14 ] }
394	480	{ "bbox": [ [ 105, 63, 165, 234 ], [ 108, 63, 180, 401 ] ], "category": [ 14, 12 ] }
500	333	{ "bbox": [ [ 142, 4, 284, 329 ] ], "category": [ 14 ] }
500	375	{ "bbox": [ [ 58, 151, 403, 224 ], [ 343, 149, 157, 226 ], [ 320, 84, 136, 278 ], [ 303, 91, 88, 118 ], [ 279, 87, 58, 85 ], [ 162, 66, ...
500	375	{ "bbox": [ [ 13, 109, 483, 247 ] ], "category": [ 6 ] }
500	375	{ "bbox": [ [ 288, 105, 50, 87 ], [ 374, 143, 29, 45 ], [ 457, 153, 23, 37 ], [ 1, 0, 311, 375 ] ], "category": [ 3, 3, 3, 3 ] }
500	375	{ "bbox": [ [ 0, 94, 240, 242 ] ], "category": [ 19 ] }
500	375	{ "bbox": [ [ 64, 38, 395, 308 ] ], "category": [ 2 ] }
500	375	{ "bbox": [ [ 425, 196, 75, 77 ], [ 234, 215, 89, 43 ], [ 140, 211, 29, 65 ], [ 50, 209, 102, 99 ], [ 0, 208, 55, 105 ], [ 313, 240, 12,...
500	375	{ "bbox": [ [ 115, 63, 241, 312 ], [ 1, 211, 48, 164 ] ], "category": [ 14, 8 ] }
500	375	{ "bbox": [ [ 79, 3, 421, 372 ], [ 0, 28, 227, 347 ], [ 1, 124, 499, 251 ] ], "category": [ 14, 14, 17 ] }
320	224	{ "bbox": [ [ 85, 35, 154, 189 ], [ 114, 18, 89, 118 ], [ 278, 76, 20, 56 ] ], "category": [ 12, 14, 14 ] }
500	335	{ "bbox": [ [ 159, 133, 127, 106 ] ], "category": [ 18 ] }
500	249	{ "bbox": [ [ 27, 18, 76, 149 ], [ 15, 53, 163, 196 ], [ 146, 22, 99, 227 ], [ 226, 18, 174, 231 ], [ 390, 12, 110, 237 ], [ 363, 211, 2...

End of preview. Expand in Data Studio

FineOWOD - OWOD open-world benchmark splits in the unified detection format

Source: PASCAL VOC 2007/2012 tars (host.robots.ox.ac.uk), COCO 2017 zips (images.cocodataset.org), and the official OWOD task lists (JosephKJ/OWOD), fetched from the vendored copy in the PROB repo (github.com/orrzohar/PROB, Apache-2.0).

Converted by the finedet project into a unified, AutoTrain-compatible layout: image / width / height / objects{bbox, category} with COCO-format [x, y, w, h] boxes in absolute pixels. Boxes are clipped to the image and empty boxes dropped; category ids are densified per the category tables below.

This repository hosts one config per source release: t1, t2 (default: t1). Load one with load_dataset("finedet/owod", "<config>").

Box format

objects.bbox follows the COCO convention: [x, y, w, h] in absolute pixels, origin at the image's top-left corner.

License

The OWOD task splits are published under Apache-2.0 (github.com/JosephKJ/OWOD). Underlying data: COCO 2017 annotations CC BY 4.0 (COCO Consortium), COCO images subject to the Flickr Terms of Use; PASCAL VOC 2007/2012 states no explicit license (images collected from Flickr).

Config `t1`

Example images

Boxes are colored by category: near-transparent fill, opaque outline.

Conversion notes

One config per OWOD task (t1-t4): train = that task's training image list, test = the common all-task test list (duplicate list lines collapsed). The 80-class vocabulary is task-ordered (ids 0-19 = t1/VOC ... 60-79 = t4) and identical across configs; per config, boxes are restricted to the classes known at that task. Task 1: boxes limited to the 20 VOC classes (dense ids 0-19). Note: the strict OWOD training protocol additionally hides the classes of earlier tasks at train time — reproduce it by dropping train boxes with category id < 20*(K-1). The protocol's 'unknown' class is not materialized.

Splits

test: 9119 images
train: 16551 images

id	name
0	aeroplane
1	bicycle
2	bird
3	boat
4	bottle
5	bus
6	car
7	cat
8	chair
9	cow
10	diningtable
11	dog
12	horse
13	motorbike
14	person
15	pottedplant
16	sheep
17	sofa
18	train
19	tvmonitor
20	truck
21	traffic light
22	fire hydrant
23	stop sign
24	parking meter
25	bench
26	elephant
27	bear
28	zebra
29	giraffe
30	backpack
31	umbrella
32	handbag
33	tie
34	suitcase
35	microwave
36	oven
37	toaster
38	sink
39	refrigerator
40	frisbee
41	skis
42	snowboard
43	sports ball
44	kite
45	baseball bat
46	baseball glove
47	skateboard
48	surfboard
49	tennis racket
50	banana
51	apple
52	sandwich
53	orange
54	broccoli
55	carrot
56	hot dog
57	pizza
58	donut
59	cake
60	bed
61	toilet
62	laptop
63	mouse
64	remote
65	keyboard
66	cell phone
67	book
68	clock
69	vase
70	scissors
71	teddy bear
72	hair drier
73	toothbrush
74	wine glass
75	cup
76	fork
77	knife
78	spoon
79	bowl

Config `t2`

Example images

Boxes are colored by category: near-transparent fill, opaque outline.

Conversion notes

One config per OWOD task (t1-t4): train = that task's training image list, test = the common all-task test list (duplicate list lines collapsed). The 80-class vocabulary is task-ordered (ids 0-19 = t1/VOC ... 60-79 = t4) and identical across configs; per config, boxes are restricted to the classes known at that task. Task 2: boxes limited to the 40 classes known at task 2 (ids 0-39). Note: the strict OWOD training protocol additionally hides the classes of earlier tasks at train time — reproduce it by dropping train boxes with category id < 20*(K-1). The protocol's 'unknown' class is not materialized.

Splits

test: 9119 images
train: 45520 images

id	name
0	aeroplane
1	bicycle
2	bird
3	boat
4	bottle
5	bus
6	car
7	cat
8	chair
9	cow
10	diningtable
11	dog
12	horse
13	motorbike
14	person
15	pottedplant
16	sheep
17	sofa
18	train
19	tvmonitor
20	truck
21	traffic light
22	fire hydrant
23	stop sign
24	parking meter
25	bench
26	elephant
27	bear
28	zebra
29	giraffe
30	backpack
31	umbrella
32	handbag
33	tie
34	suitcase
35	microwave
36	oven
37	toaster
38	sink
39	refrigerator
40	frisbee
41	skis
42	snowboard
43	sports ball
44	kite
45	baseball bat
46	baseball glove
47	skateboard
48	surfboard
49	tennis racket
50	banana
51	apple
52	sandwich
53	orange
54	broccoli
55	carrot
56	hot dog
57	pizza
58	donut
59	cake
60	bed
61	toilet
62	laptop
63	mouse
64	remote
65	keyboard
66	cell phone
67	book
68	clock
69	vase
70	scissors
71	teddy bear
72	hair drier
73	toothbrush
74	wine glass
75	cup
76	fork
77	knife
78	spoon
79	bowl

Training with transformers

The boxes are already in the absolute-pixel COCO [x, y, w, h] format that AutoImageProcessor expects, so fine-tuning a detector needs no bbox conversion:

import torch
from datasets import load_dataset
from transformers import (AutoImageProcessor, AutoModelForObjectDetection,
                          Trainer, TrainingArguments)

ds = load_dataset("finedet/owod", "t1")
obj_feat = ds["train"].features["objects"]
if hasattr(obj_feat, "feature"):
    obj_feat = obj_feat.feature
cat_feat = obj_feat["category"]
names = (cat_feat.feature if hasattr(cat_feat, "feature") else cat_feat).names

checkpoint = "facebook/detr-resnet-50"
processor = AutoImageProcessor.from_pretrained(checkpoint)
model = AutoModelForObjectDetection.from_pretrained(
    checkpoint,
    id2label=dict(enumerate(names)),
    label2id={n: i for i, n in enumerate(names)},
    ignore_mismatched_sizes=True,
)


def transform(batch):
    images = [img.convert("RGB") for img in batch["image"]]
    annotations = [
        {"image_id": i,
         "annotations": [
             {"bbox": box, "category_id": cat, "area": box[2] * box[3], "iscrowd": 0}
             for box, cat in zip(objs["bbox"], objs["category"])
         ]}
        for i, objs in enumerate(batch["objects"])
    ]
    return processor(images=images, annotations=annotations, return_tensors="pt")


def collate(batch):
    return {"pixel_values": torch.stack([x["pixel_values"] for x in batch]),
            "labels": [x["labels"] for x in batch]}


trainer = Trainer(
    model=model,
    args=TrainingArguments(output_dir="out", per_device_train_batch_size=4,
                           num_train_epochs=10, learning_rate=1e-5,
                           remove_unused_columns=False),
    train_dataset=ds["train"].with_transform(transform),
    data_collator=collate,
)
trainer.train()

Downloads last month: 73

Datasets:

finedet
/

owod

FineOWOD - OWOD open-world benchmark splits in the unified detection format

Box format

License

Config `t1`

Example images

Conversion notes

Splits

Categories

Config `t2`

Example images

Conversion notes

Splits

Categories

Training with transformers

FineOWOD - OWOD open-world benchmark splits in the unified detection format

Box format

License

Config t1

Example images

Conversion notes

Splits

Categories

Config t2

Example images

Conversion notes

Splits

Categories

Training with transformers

Config `t1`

Config `t2`