YouTube Skijump Dataset

This dataset consists of images from 10 publicly available videos of skijump competitions on YouTube. It contains keypoint annotations made by hand and some segmentation masks that are only partly correct and mostly automatically retrieved. The dataset contains 2867 annotated images and is split into train/test/val subsets with 2159, 560 and 148 images, resprectively. The train and validation annotations are included in the download package. If you are interested in the test annotations as well, contact  Katja Ludwig via E-Mail.


The annotation files of the dataset include the following information:

  • The name of the athlete, retrieved from the video footage.
  • A flag indicating if the image belongs to a slowmotion replay or not.
  • 2D image keypoint coordinates from head, r. shoulder, r. elbow, r. hand, l. shoulder, l. elbow, l. hand, r. hip, r. knee, r. ankle, l. hip, l. knee, l. ankle, r. ski tip, r. ski tail, l. ski tip, l. ski tail and their visibility (0: invisible, 1: occluded, 2: visible)

The dataset splits are created such that each athlete is only contained in one split.


Segmentation masks

The dataset further contains 424 segmentation masks. Most of them are automatically generated as described in our paper. These segmentation masks are only partly correct. This means that the segmentations for some body parts are wrong, incomplete or missing. The following body parts are included in the segmentation masks: head, torso, l./r. upper arm, l./r. forearm, l./r. hand, l./r. thigh, l./r. lower leg, l./r. foot, l./r. ski


If you are using this dataset, please cite the following paper

    title     = {Detecting Arbitrary Keypoints on Limbs and Skis with Sparse Partly Correct Segmentation Masks},
    author    = {Ludwig, Katja and Kienzle, Daniel and Lorenz, Julian and Lienhart, Rainer},
    booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
    month     = {January},
    year      = {2023},



You can download the dataset here:


If you have any questions or want access to the test set annotations, contact  Katja Ludwig



These files provide 3515 left-right annotation corrections for the DensePose subset of the MSCOCO-Dataset. Some of these corrections were made per hand, others according to heuristics. Therefore, the dataset might still contain some small errors. You will find corrections for upper arm, lower arm, thigh and lower leg, other bodyparts were not considered. The indices in the files correspond to the following standard bodypart indices used in the DensePose data:


1: Torso, 2: Right Hand, 3: Left Hand, 4: Left Foot, 5: Right Foot, 6: Upper Leg Right, 7: Upper Leg Left, 8: Lower Leg Right,  9: Lower Leg Left, 10: Upper Arm Left, 11: Upper Arm Right, 12: Lower Arm Left, 13: Lower Arm Right, 14: Head


The files contain the MSCOCO image id, the person number (in the order of the MSCOCO annotations in the original *.json files, 0-indexed) and the ids of the body parts to be swapped according to the list.







If you have any questions, contact  Katja Ludwig



© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg
© Universität Augsburg

The FlickrLogos dataset consists of real-world images collected from Flickr depicting company logos in various circumstances. The dataset comes in two versions: The original FlickrLogos-32 dataset and the FlickrLogos-47 dataset. FlickrLogos-32 was designed for logo retrieval and multi-class logo detection and object recognition. However, the annotations for object detection were often incomplete,since only the most prominent logo instances were labelled. FlickrLogos-47 uses the same image corpus as FlickrLogos-32 but has been re-annotated specifically for the task of object detection and recognition. New classes were introduced (i.e. the company logo and text are now treated as separate classes where applicable) and missing object instances have been annotated. Furthermore, we provide corresponsing evaluation scripts.



The dataset FlickrLogos-32 contains photos showing brand logos and is meant for the evaluation of logo retrieval and multi-class logo detection/recognition systems on real-world images. We collected logos of 32 different logo brands by downloading them from Flickr. All logos have an approximately planar surface.



There are 32 logo classes: Adidas, Aldi, Apple, Becks, BMW, Carlsberg, Chimay, Coca-Cola, Corona, DHL, Erdinger, Esso, Fedex, Ferrari, Ford, Foster's, Google, Guiness, Heineken, HP, Milka, Nvidia, Paulaner, Pepsi, Ritter Sport, Shell, Singha, Starbucks, Stella Artois, Texaco, Tsingtao and UPS.


Partitions / subsets:

The retrieved images were inspected manually to ensure that the specific logo is actually shown. The whole dataset is split into three disjoint subsets P1, P2, and P3, each containing images of all 32 classes. The first partition P1 - the training set - consists of 10 images that were hand-picked such that these consistently show a single logo under various views with as little background clutter as possible. The other two partitions P2 (validation set) and P3 (test set = query set) contain 30 images per class. Unlike P1 these images contain at least one instance of a logo but in several cases multiple instances.


To facilitate the development of high-precision classifiers the evaluation of their sensitivity on non-logo images is important. Therefore both partitions P2, and P3 include another 3000 images downloaded from Flickr with the queries "building", "nature", "people" and "friends". These images are the negative images and complete our dataset. A brief summary of the data subsets is shown in the table below.


Dataset Partitions / Subsets
Partition Description Images #Images
P1 (training set)

Hand-picked images

10 per class 320 images
P2 (validation set) Images showing at least a single logo under various views 30 per class 3960 images
Non-logo images 3000
P3 (test set = query set) Images showing at least a single logo under various views 30 per class 3960 images
Non-logo images 3000
P1, P2 and P3 are disjoint.   8240 images


description here description here
description here description here
description here description here


Pixel-level annotations

This dataset further includes pixel-level annotations, i.e. binary masks + bounding boxes (see above) that mark the position of the logo in each image.

The binary masks are provided as .png files as well as by the coordinates of the corresponding bounding box.
Each logo image in <basedir>/classes/jpg/<class>/<file>.jpg has its mask in <basedir>/classes/masks/<class>/<file>.mask.<n>.png where n is the mask number starting at 0.
Masks are single-channel PNG images of the same size as the original image. Masked areas (where the logos is) have a value != 0, unmasked areas (=background) 0.

There are separate masks for each annotation in an image available and an additional mask merged from all individual masks:
Each logo image in <basedir>/classes/jpg/<class>/<file>.jpg has its merged mask in <basedir>/classes/masks/<class>/<file>.mask.merged.png.

Also thumbnail images are included for easy visualization of results.

Evaluation Protocol

Evaluation of classification/recognition methods:
  • Training may use training + validation set only.
  • All the images in the test set P3 (logos+non-logos = 3960 images) are used for the evaluation of recognition methods.

Evaluation of retrieval methods:

  • All images in training and validation set are indexed, including non-logo images = 4280 images in total.
  • All images in query set P3 (logos only = 960 images) are used for the evaluation of retrieval methods.
  • The average precision is computed for every query and averaged over all queries yielding the final mAP.
  • If not specified otherwise the masks/bounding boxes are not used as region-of-interest when indexing and querying.

Evaluation & Tools

We provide evaluation scripts to test both retrieval and classification systems (Python 2.7+ needed):
  • Evaluates the retrieval of logo images by computing mAP, AvgTop4 score, response ratio, etc.
  • Evaluates the classification of logo images by computing precision, recall and more.
  • Plots classification results, i.e. true positives per class and the confusion matrix. Contained in full package.
  • Other scripts that simplify the management of this dataset (copying, etc..)
Download the evaluation kit including all scripts plus additional sample data. The evaluation kit was last updated on 18th November 2013 (Version 1.0.4). If you encounter problems, please contact us.


If you use this dataset in your work please cite the following paper:

[1] Scalable Logo Recognition in Real-World Images
Stefan Romberg, Lluis Garcia Pueyo, Rainer Lienhart, Roelof van Zwol
ACM International Conference on Multimedia Retrieval 2011 (ICMR11), Trento, April 2011.
Also Technical Report, University of Augsburg, Institute of Computer Science, March 2011



The dataset FlickrLogos-47 contains photos showing brand logos and is meant for the evaluation of logo detection and recognition systems on real-world images. It consists of the same images as in the FlickrLogos-32 dataset, but has been re-annotated to fix missing annotations and to include more classes.



There are 47 logo classes: Adidas (Symbol), Adidas (Text), Aldi, Apple, Becks (Symbol), Becks (Text), BMW, Carlsberg (Symbol), Carlsberg (Text), Chimay (Symbol), Chimay (Text), Coca-Cola, Corona (Symbol), Corona (Text), DHL, Erdinger (Symbol), Erdinger (Text), Esso (Symbol), Esso (Text), Fedex, Ferrari, Ford, Foster's (Symbol), Foster's (Text), Google, Guiness (Symbol), Guiness (Text), Heineken, HP, Milka (Symbol), Milka (Text), Nvidia (Symbol), Nvidia (Text), Paulaner (Symbol), Paulaner (Text), Pepsi (Symbol), Pepsi (Text), Ritter Sport, Shell, Singha (Symbol), Singha (Text), Starbucks, Stella Artois (Symbol), Stella Artois (Text), Texaco, Tsingtao (Symbol) Tsingtao (Text) and UPS.




If you wish to download one (or both of the datasets), please send an (informal) email to Please state your name, your institution and why you would like to have access to this dataset (we are curious). We then send you a download link by e-mail.

Note: This dataset consists of images downloaded from Flickr. Use of these images must respect Flickr's terms of use.


Important Notes

There is a similar dataset called FlickrLogos-27. It consists of 27 classes only and while there is some overlap it is largely different from our FlickrLogos-32 dataset. Results for these different datasets are not comparable.



If you have any questions, corrections or other issues please contact Stephan Brehm. The former maintainers were Christian Eggert and Stefan Romberg.