Jointly Discovering Visual Objects and Spoken Words from Raw Sensory Input

TL;DR

This paper presents neural networks that learn to associate spoken words with visual objects directly from raw audio and image data without explicit supervision, enabling emergent object and word detectors.

Contribution

It introduces a method for unsupervised learning of audio-visual associations that discovers objects and spoken words simultaneously from raw sensory inputs.

Findings

Models learn semantically-coupled object and word detectors

Associations emerge without explicit labels or segmentations

Effective on Places 205 and ADE20k datasets

Abstract

In this paper, we explore neural network models that learn to associate segments of spoken audio captions with the semantically relevant portions of natural images that they refer to. We demonstrate that these audio-visual associative localizations emerge from network-internal representations learned as a by-product of training to perform an image-audio retrieval task. Our models operate directly on the image pixels and speech waveform, and do not rely on any conventional supervision in the form of labels, segmentations, or alignments between the modalities during training. We perform analysis using the Places 205 and ADE20k datasets demonstrating that our models implicitly learn semantically-coupled object and word detectors.