All You Can Embed: Natural Language based Vehicle Retrieval with Spatio-Temporal Transformers

TL;DR

This paper introduces AYCE, a modular model that combines natural language and visual data using transformers for vehicle retrieval in smart cities, demonstrating effective cross-modal embedding techniques.

Contribution

The paper presents a novel architecture integrating BERT and transformers for vehicle retrieval, with a new triplet loss variation for cross-modal embedding alignment.

Findings

Effective vehicle retrieval using combined visual and textual data

Modular architecture allows flexible integration of language and vision

Publicly available code facilitates reproducibility and further research

Abstract

Combining Natural Language with Vision represents a unique and interesting challenge in the domain of Artificial Intelligence. The AI City Challenge Track 5 for Natural Language-Based Vehicle Retrieval focuses on the problem of combining visual and textual information, applied to a smart-city use case. In this paper, we present All You Can Embed (AYCE), a modular solution to correlate single-vehicle tracking sequences with natural language. The main building blocks of the proposed architecture are (i) BERT to provide an embedding of the textual descriptions, (ii) a convolutional backbone along with a Transformer model to embed the visual information. For the training of the retrieval model, a variation of the Triplet Margin Loss is proposed to learn a distance measure between the visual and language embeddings. The code is publicly available at https://github.com/cscribano/AYCE_2021.