Written by 8:57 pm Deep Learning 2 komentarze

Open Images Dataset V4

Open Images Dataset V4 - Image examples

30th April 2018 new version of Open Images Dataset V4 is released. There is also announced a challenge for best object detection results using this dataset.

Here you can see data examples: Open Images Dataset V4

ECCV 2018 Open Images Challenge

During ECCV 2018 conference there will be a workshop dedicated  Open Images Challenge (presented by Vittorio Ferrari, Alina Kuznetsova, Jasper Uijlings, Rodrigo Benenson, Victor Gomes, Matteo Malloci). They will announce challenge results.

The Challenge has two tracks:

  1. Object Class Detection: predicting a tight bounding box around all instances of the 500 classes.
  2. Visual Relationship Detection: detecting pairs of objects in particular relations, e.g. „woman playing guitar”.


There are image-level labels (table 1) and bounding boxes for object detection (table 2). Labels where generated automatically by machine and then checked by humans thanks to Crowdsource labeler.

Table 1: Image-level labels

Train Validation Test # Classes # Trainable Classes
Images 9,011,219 41,620 125,436
Machine-Generated Labels 78,977,695 512,093 1,545,835 7,870 4,764
Human-Verified Labels 27,894,289
pos:   13,444,569
neg: 14,449,720
pos: 365,772
neg: 185,618
pos: 1,105,052
neg:    562,347
19,794 7,186

Table 2: Object Detection track annotations on training set

Classes Images Image-Level Labels Bounding boxes
Train 500 1,743,042 5,743,460
pos:   3,830,005
neg: 1,913,455


Here are some examples, more you can find here: Open Images Dataset V4


  • April 30th 2018: training set for object detection track released (with bounding box annotations).
  • May 10 2018: visual relationship detection annotations on the training set will be released.
  • May 31 2018: evaluation metric protocols and implementation will be released (as a part of the TF Object Detection API).
  • July 1st 2018: a test set of 100k images will be released by Kaggle.
  • September 1st 2018: deadline for submission of results.
  • September 8th 2018: Open Images Challenge Workshop