We invite researchers to participate in a challenge for large-scale scene understanding models as part of the LSUN workshop at this year’s Computer Vision and Pattern Recognition (CVPR) conference.

Mapillary is co-organizing the Large-Scale Scene Understanding (LSUN) workshop at this year’s CVPR conference in Honolulu, Hawaii. This is the premier annual conference on Computer Vision and Pattern Recognition. The LSUN workshop brings together researchers working on novel object recognition models for interpreting what is going on in image scenes.

LSUN involves several tasks for scene understanding. Mapillary is responsible for two of those: Semantic Image Segmentation and Instance-specific Image Segmentation of Street-level Images. The workshop offers researchers a great opportunity to challenge themselves and the segmentation models they have been working on.

In the Semantic Image Segmentation task, the goal is to label each pixel in an image with an object category from a predefined set. The Instance-specific Image Segmentation task requires that each car, pedestrian, etc. is segmented and labelled individually. Please find the detailed rules for each task on our research page.

Semantic Image Segmentation Task
Instance-specific Semantic Segmentation Task

Both tasks will be using the Mapillary Vistas Dataset that we recently released. The Research edition of the dataset contains 66 annotated object classes, 37 of them instance-specific, covering street-level scenes from all around the world. Today it is the largest and most diverse dataset of its kind that is publicly available. Read more about the dataset, its quality assessment, and how to access it.

To join the challenge, you need to submit your segmentation results to our benchmark server by July 9 according to the instructions here. No results or leaderboards will be published during the challenge. The winner and final leaderboard will be announced during the LSUN workshop at CVPR on July 26.

We are looking forward to your participation in the challenge, and please spread the word to anyone you think might be interested in this. As always, we’re here for any questions and comments via research@mapillary.com.

/Peter and Mapillary Research

Tags for this post: challenges computervision
comments powered by Disqus