# Large-Scale Object Mining for Object Discovery from Unlabeled Video

## TL;DR

This paper presents a method for automatic object discovery in unlabeled automotive videos by mining object tracks, and introduces a large dataset of over 360,000 object tracks to facilitate further research.

## Contribution

It demonstrates the feasibility of fully automatic object discovery in automotive videos and provides a new large-scale dataset for evaluating feature representations and clustering strategies.

## Key findings

- Feasibility of automatic object discovery in unlabeled videos
- Introduction of a dataset with over 360,000 object tracks
- Evaluation of feature and clustering methods for object discovery

## Abstract

This paper addresses the problem of object discovery from unlabeled driving videos captured in a realistic automotive setting. Identifying recurring object categories in such raw video streams is a very challenging problem. Not only do object candidates first have to be localized in the input images, but many interesting object categories occur relatively infrequently. Object discovery will therefore have to deal with the difficulties of operating in the long tail of the object distribution. We demonstrate the feasibility of performing fully automatic object discovery in such a setting by mining object tracks using a generic object tracker. In order to facilitate further research in object discovery, we release a collection of more than 360,000 automatically mined object tracks from 10+ hours of video data (560,000 frames). We use this dataset to evaluate the suitability of different feature representations and clustering strategies for object discovery.