ViCo: Word Embeddings from Visual Co-occurrences
This addresses the challenge of representing visual concepts in word embeddings for vision-language applications, though it is incremental as it builds on existing methods.
The authors tackled the problem of learning word embeddings from visual co-occurrences in images, showing that these embeddings complement text-only ones like GloVe by better capturing visual concept similarities, and achieved gains on five downstream vision-language tasks when augmenting GloVe.
We propose to learn word embeddings from visual co-occurrences. Two words co-occur visually if both words apply to the same image or image region. Specifically, we extract four types of visual co-occurrences between object and attribute words from large-scale, textually-annotated visual databases like VisualGenome and ImageNet. We then train a multi-task log-bilinear model that compactly encodes word "meanings" represented by each co-occurrence type into a single visual word-vector. Through unsupervised clustering, supervised partitioning, and a zero-shot-like generalization analysis we show that our word embeddings complement text-only embeddings like GloVe by better representing similarities and differences between visual concepts that are difficult to obtain from text corpora alone. We further evaluate our embeddings on five downstream applications, four of which are vision-language tasks. Augmenting GloVe with our embeddings yields gains on all tasks. We also find that random embeddings perform comparably to learned embeddings on all supervised vision-language tasks, contrary to conventional wisdom.