Dataset Ninja LogoDataset Ninja:

LogoDet-3K Dataset

15865430001793
Tagretail
Taskobject detection
Release YearMade in 2020
Licenseunknown

Introduction #

Released 2020-08-12 Β·Jing Wang, Weiqing Min, Sujuan Houet al.

The LogoDet-3K: A Large-Scale Image Dataset for Logo Detection is a significant contribution to the logo detection, with broad applications in multimedia such as copyright infringement detection, brand visibility monitoring, and product brand management on social media. It features 3,000 logo categories, approximately 200,000 manually annotated logo objects, and 158,652 images, sets a new standard for logo detection benchmarks. The authors detail the dataset construction process, including logo image collection, filtering, and object annotation. The dataset is divided into nine super-classes based on daily life needs and common enterprise positioning.

For the construction of LogoDet-3K, the authors followed a meticulous three-step process, involving logo image collection, logo image filtering, and logo object annotation. Each image was thoroughly examined and reviewed to ensure the dataset’s quality after filtering and annotation.

Logo image collection

During the phase, it was compiled a comprehensive logo list based on various reputable sources, resulting in a vocabulary of 3,000 logo names spanning nine super-classes: sports, leisure, transportation, food, electronic, necessities, clothes, medical, others. Images were then retrieved from search engines using these names as queries. The dataset creation process involved manual cleaning and filtering to ensure data quality, including the removal of images with unsuitable sizes or extreme aspect ratios and those without logos.

Root Category Sub-Category Images Objects
Food 932 53,350 64,276
Clothes 604 31,266 37,601
Necessities 432 24,822 30,643
Others 371 15,513 20,016
Electronic 224 9,675 12,139
Transportation 213 10,445 12,791
Leisure 111 5,685 6,573
Sports 66 3,945 5,041
Medical 47 3,945 5,185
Total 3,000 158,652 194,261

Logo object annotation

During the process a considerable amount of time was required. The resulting LogoDet-3K dataset consists of 3,000 logo classes, 158,652 images, and 194,261 logo objects. The dataset exhibits imbalanced distributions across different logo categories, presenting a challenge for effective logo detection, particularly for categories with fewer samples. The statistics at the superclass and category levels illustrate this imbalance.

image

Multiple logo categories for some brands, where a distinction between these logo categories via adding the suffix β€˜-1’, β€˜-2’.

Additionally, the authors provide insights into the distribution of images and categories in LogoDet-3K, highlighting the challenges posed by imbalances across different logo objects and images. The dataset is characterized by a large percentage of small and medium logo objects, creating additional challenges for logo detection, as smaller logos are inherently harder to detect.

image

Sorted distribution of images for each logo in LogoDet-3K.

The dataset’s statistics further detail the distribution of logo categories, images, and logo objects across nine super-classes, revealing variations in sizes and numbers. Notably, the Food, Clothes, and Necessities classes exhibit larger numbers of objects and images compared to other classes in the dataset.

image

The detailed statistics of LogoDet-3K about Image and object distribution in per category, the number of objects in per image and object size in per image.

image

Distributions of categories, images and objects from LogoDet-3K on super-classes.

ExpandExpand
Dataset LinkHomepageDataset LinkResearch PaperDataset LinkKaggle

Summary #

LogoDet-3K is a dataset for an object detection task. It is used in the retail industry.

The dataset consists of 158654 images with 194261 labeled objects belonging to 3000 different classes including lexus-1, avery_dennison-2, alpinestars-2, and other: new_balance-1, 76, maybach-1, mclaren, lexus-2, kichesippi-2, zendium, yuyue-2, nongfu_spring-2, longmu-1, longmu-2, square_one_organic, steeden, neil_pryde-2, new_balance-2, tiffany, kichesippi-1, yuyue-1, auxx-2, chow_tai_seng-1, thomapyrin, kleenex, jus-rol, paralen, music_man-2, and 2972 more.

Images in the LogoDet-3K dataset have bounding box annotations. There are 4 (0% of the total) unlabeled images (i.e. without annotations). There are no pre-defined train/val/test splits in the dataset. Alternatively, the dataset could be split into 9 root-categories: food (53350 images), clothes (31266 images), necessities (24822 images), others (15513 images), transportation (10445 images), electronic (9675 images), leisure (5685 images), sports (3953 images), and medical (3945 images). The dataset was released in 2020 by the Shandong Normal University, China and Chinese Academy of Sciences.

Here is a visualized example for randomly selected sample classes:

Explore #

LogoDet-3K dataset has 158654 images. Click on one of the examples below or open "Explore" tool anytime you need to view dataset images with annotations. This tool has extended visualization capabilities like zoom, translation, objects table, custom filters and more. Hover the mouse over the images to hide or show annotations.

OpenSample annotation mask from LogoDet-3KSample image from LogoDet-3K
OpenSample annotation mask from LogoDet-3KSample image from LogoDet-3K
OpenSample annotation mask from LogoDet-3KSample image from LogoDet-3K
OpenSample annotation mask from LogoDet-3KSample image from LogoDet-3K
OpenSample annotation mask from LogoDet-3KSample image from LogoDet-3K
OpenSample annotation mask from LogoDet-3KSample image from LogoDet-3K
OpenSample annotation mask from LogoDet-3KSample image from LogoDet-3K
OpenSample annotation mask from LogoDet-3KSample image from LogoDet-3K
OpenSample annotation mask from LogoDet-3KSample image from LogoDet-3K
OpenSample annotation mask from LogoDet-3KSample image from LogoDet-3K
OpenSample annotation mask from LogoDet-3KSample image from LogoDet-3K
OpenSample annotation mask from LogoDet-3KSample image from LogoDet-3K
πŸ‘€
Have a look at 158654 images
Because of dataset's license preview is limited to 12 images
View images along with annotations and tags, search and filter by various parameters

Class balance #

There are 3000 annotation classes in the dataset. Find the general statistics and balances for every class in the table below. Click any row to preview images that have labels of the selected class. Sort by column to find the most rare or prevalent classes.

Search
Rows 1-10 of 3000
Class
γ…€
Images
γ…€
Objects
γ…€
Count on image
average
Area on image
average
lexus-1βž”
rectangle
426
466
1.09
28.49%
avery_dennison-2βž”
rectangle
352
532
1.51
17.94%
alpinestars-2βž”
rectangle
328
508
1.55
34.77%
new_balance-1βž”
rectangle
308
328
1.06
30.82%
76βž”
rectangle
302
302
1
15.68%
maybach-1βž”
rectangle
294
310
1.05
32.92%
mclarenβž”
rectangle
252
258
1.02
22.95%
kichesippi-2βž”
rectangle
250
260
1.04
10%
lexus-2βž”
rectangle
250
270
1.08
18.27%
zendiumβž”
rectangle
249
312
1.25
11.55%

Images #

Explore every single image in the dataset with respect to the number of annotations of each class it has. Click a row to preview selected image. Sort by any column to find anomalies and edge cases. Use horizontal scroll if the table has many columns for a large number of classes in the dataset.

Class sizes #

The table below gives various size properties of objects for every class. Click a row to see the image with annotations of the selected class. Sort columns to find classes with the smallest or largest objects or understand the size differences between classes.

Search
Rows 1-10 of 2993
Class
Object count
Avg area
Max area
Min area
Min height
Min height
Max height
Max height
Avg height
Avg height
Min width
Min width
Max width
Max width
avery_dennison-2
rectangle
532
11.88%
44.36%
1.17%
31px
7.93%
390px
99.74%
119px
31.53%
46px
9.02%
385px
76.85%
alpinestars-2
rectangle
508
23.29%
71.28%
2.93%
30px
7.53%
393px
98.95%
121px
31.75%
105px
27.48%
520px
99.81%
lexus-1
rectangle
466
26.04%
88.39%
0.71%
38px
10.03%
384px
98.97%
180px
47.21%
32px
6.56%
512px
99.71%
violet_crumble
rectangle
414
12.17%
79.47%
0.72%
16px
4.21%
315px
88.73%
90px
21.85%
26px
5.3%
517px
99.81%
new_balance-1
rectangle
328
28.94%
99.55%
1.57%
27px
7.18%
395px
99.75%
176px
46.22%
81px
16.14%
510px
99.8%
auxx-2
rectangle
327
6.92%
52.51%
0.43%
18px
4.31%
349px
99.71%
72px
18.89%
46px
9%
475px
97.74%
biggby_coffee
rectangle
320
7.64%
88.34%
0.25%
16px
3.52%
377px
98.43%
101px
26.01%
16px
3.07%
464px
89.75%
zendium
rectangle
312
9.23%
64.68%
1.13%
26px
5.5%
308px
69.25%
99px
25.32%
22px
4.35%
509px
99.8%
bacardi
rectangle
312
18.81%
98.95%
1.01%
29px
7.27%
424px
99.74%
138px
36.09%
55px
13.92%
504px
99.8%
maybach-1
rectangle
310
31.22%
85.7%
2.26%
69px
17.6%
376px
96.63%
201px
52.22%
59px
11.73%
507px
98.64%

Spatial Heatmap #

The heatmaps below give the spatial distributions of all objects for every class. These visualizations provide insights into the most probable and rare object locations on the image. It helps analyze objects' placements in a dataset.

Spatial Heatmap

Objects #

Table contains all 100205 objects. Click a row to preview an image with annotations, and use search or pagination to navigate. Sort columns to find outliers in the dataset.

Search
Rows 1-10 of 100205
Object ID
γ…€
Class
γ…€
Image name
click row to open
Image size
height x width
Height
γ…€
Height
γ…€
Width
γ…€
Width
γ…€
Area
γ…€
1βž”
embraer
rectangle
Embraer_46.jpg
350 x 495
174px
49.71%
317px
64.04%
31.84%
2βž”
cafe_du_monde
rectangle
Cafe du Monde_122.jpg
373 x 489
63px
16.89%
195px
39.88%
6.74%
3βž”
ermenegildo_zegna
rectangle
Ermenegildo Zegna_42.jpg
378 x 484
97px
25.66%
462px
95.45%
24.49%
4βž”
taixiang
rectangle
taixiang_23.jpg
397 x 521
83px
20.91%
136px
26.1%
5.46%
5βž”
pakistan_state
rectangle
pakistan state_28.jpg
362 x 489
249px
68.78%
230px
47.03%
32.35%
6βž”
pakistan_state
rectangle
pakistan state_28.jpg
362 x 489
74px
20.44%
271px
55.42%
11.33%
7βž”
timex
rectangle
timex_142.jpg
389 x 517
186px
47.81%
286px
55.32%
26.45%
8βž”
count_chocula
rectangle
Count Chocula_83.jpg
524 x 361
124px
23.66%
184px
50.97%
12.06%
9βž”
laura_secord_chocolates
rectangle
laura secord chocolates_67.jpg
378 x 519
52px
13.76%
202px
38.92%
5.35%
10βž”
gigo
rectangle
gigo_46.jpg
380 x 487
55px
14.47%
72px
14.78%
2.14%

License #

License is unknown for the LogoDet-3K dataset.

Source

Citation #

If you make use of the LogoDet-3K data, please cite the following reference:

@dataset{LogoDet-3K,
  author={Jing Wang and Weiqing Min and Sujuan Hou and Shengnan Ma and Yuanjie Zheng and Shuqiang Jiang},
  title={LogoDet-3K},
  year={2020},
  url={https://github.com/Wangjing1551/LogoDet-3K-Dataset}
}

Source

If you are happy with Dataset Ninja and use provided visualizations and tools in your work, please cite us:

@misc{ visualization-tools-for-logodet-3k-dataset,
  title = { Visualization Tools for LogoDet-3K Dataset },
  type = { Computer Vision Tools },
  author = { Dataset Ninja },
  howpublished = { \url{ https://datasetninja.com/logodet-3k } },
  url = { https://datasetninja.com/logodet-3k },
  journal = { Dataset Ninja },
  publisher = { Dataset Ninja },
  year = { 2024 },
  month = { jul },
  note = { visited on 2024-07-25 },
}

Download #

Please visit dataset homepage to download the data.

. . .

Disclaimer #

Our gal from the legal dep told us we need to post this:

Dataset Ninja provides visualizations and statistics for some datasets that can be found online and can be downloaded by general audience. Dataset Ninja is not a dataset hosting platform and can only be used for informational purposes. The platform does not claim any rights for the original content, including images, videos, annotations and descriptions. Joint publishing is prohibited.

You take full responsibility when you use datasets presented at Dataset Ninja, as well as other information, including visualizations and statistics we provide. You are in charge of compliance with any dataset license and all other permissions. You are required to navigate datasets homepage and make sure that you can use it. In case of any questions, get in touch with us at hello@datasetninja.com.