Introduction
[Google AI](https://ai.google/)
(Google’s AI research arm, tasked with advancing AI for everyone) is challenging you to build an algorithm that detects objects automatically using an absolutely massive training dataset ― one with more varied and complex bounding-box annotations and object classes than ever before.
Here's the background. Computers are getting better and better at vision. But in a few critical ways, they still can't match a human’s intuitive perception.
For example, what do you see when you look at this photo?
![1](https://storage.googleapis.com/kaggle-media/competitions/google-research/1.png)
Most of us would answer, “a sandy beach, the ocean, a few people walking, some trees, grass, and buildings…a woman walking her dog right there! Oh yeah, and there is a man holding a plastic cup.”
Can a computer provide as precise an image description? Google AI wants to further push the capabilities of computer vision. We hope that providing very large training set will stimulate research into more sophisticated object and relationship detection models that will exceed current state-of-the-art performance.
The results of this Challenge will be presented at a workshop at the
[European Conference on Computer Vision 2018.](https://eccv2018.org/)
[Visual Relationship Detection Track Identifying different objects (man and cup) is an important problem on its own, but identifying the relationship between them (holding) is critical for many real world use cases. In this Visual Relationship Detection track Challenge you’re asked to build an algorithm that detects pairs of objects in particular relations: things like "woman playing guitar," "beer on table," or "dog inside car." The Challenge dataset includes both object bounding boxes and visual relationship annotations. The training set contains annotations for 329 distinct relationship triplets, occurring a total of 374,768 times.](https://eccv2018.org/)
[In this track of the Challenge, you are asked to build the best performing algorithm for automatically detecting relationships triplets.  Please refer to the](https://eccv2018.org/)
[Open Images Challenge](https://storage.googleapis.com/openimages/web/challenge.html)
page for additional details on the dataset.
This competition is one of two tracks in the Open Images Challenge.  Find the Object Detection track of this competition using the entire training set
[here](https://www.kaggle.com/c/google-ai-open-images-object-detection-track)
.
![](https://storage.googleapis.com/openimages/web/images/30e167061b6dd834.png)
Example of ‘man playing guitar’
[Radiofiera - Villa Cordellina Lombardi, Montecchio Maggiore (VI) - agosto 2010](https://www.flickr.com/photos/tomjoad/4938460850)
by
[Andrea Sartorati](https://www.flickr.com/photos/tomjoad/)
![](https://storage.googleapis.com/openimages/web/images/6dad067898615733.png)
Example of ‘chair at table’
[Epic Fireworks - Loads A Room](https://www.flickr.com/photos/epicfireworks/4843249505)
by
[Epic Fireworks](https://www.flickr.com/photos/epicfireworks/)