Towards Object Detection from Motion

TL;DR

This paper introduces a weakly supervised object detection method that learns to detect objects from motion cues using only two videos, eliminating the need for annotated images.

Contribution

It proposes a novel approach that trains object detectors based on physical plausibility of motion from videos without explicit annotations.

Findings

Successfully detects objects in robotics scenarios

Learns from motion cues without object location labels

Performs well in various robotics settings

Abstract

We present a novel approach to weakly supervised object detection. Instead of annotated images, our method only requires two short videos to learn to detect a new object: 1) a video of a moving object and 2) one or more "negative" videos of the scene without the object. The key idea of our algorithm is to train the object detector to produce physically plausible object motion when applied to the first video and to not detect anything in the second video. With this approach, our method learns to locate objects without any object location annotations. Once the model is trained, it performs object detection on single images. We evaluate our method in three robotics settings that afford learning objects from motion: observing moving objects, watching demonstrations of object manipulation, and physically interacting with objects (see a video summary at https://youtu.be/BH0Hv3zZG_4).