Unifying Visual-Semantic Embeddings with Multimodal Neural Language Models

TLDR

Our pipeline effectively unifies joint image‑text embedding models with multimodal neural language models. Inspired by recent advances in multimodal learning and machine translation, we introduce an encoder‑decoder pipeline that learns a multimodal joint embedding space for images and text and a novel language model for decoding distributed representations from that space. The encoder‑decoder pipeline uses a structure‑content neural language model that disentangles sentence structure from content conditioned on encoder representations, enabling image‑sentence ranking and novel caption generation from scratch, with sample captions for 800 images provided for comparison. Using LSTM‑based sentence encoding, the model matches state‑of‑the‑art performance on Flickr8K and Flickr30K without object detections, achieves new best results with a 19‑layer Oxford convolutional network, and demonstrates that linear encoders capture multimodal regularities, as shown by vector‑arithmetic examples such as transforming a blue car image to a red car.

Abstract

Inspired by recent advances in multimodal learning and machine translation, we introduce an encoder-decoder pipeline that learns (a): a multimodal joint embedding space with images and text and (b): a novel language model for decoding distributed representations from our space. Our pipeline effectively unifies joint image-text embedding models with multimodal neural language models. We introduce the structure-content neural language model that disentangles the structure of a sentence to its content, conditioned on representations produced by the encoder. The encoder allows one to rank images and sentences while the decoder can generate novel descriptions from scratch. Using LSTM to encode sentences, we match the state-of-the-art performance on Flickr8K and Flickr30K without using object detections. We also set new best results when using the 19-layer Oxford convolutional network. Furthermore we show that with linear encoders, the learned embedding space captures multimodal regularities in terms of vector space arithmetic e.g. *image of a blue car* - "blue" + "red" is near images of red cars. Sample captions generated for 800 images are made available for comparison.