Look Before you Speak: Visually Contextualized Utterances
This work provides a method for training visually contextualized conversational AI without manual annotations, which is significant for researchers and developers in multimodal AI and conversational systems.
This paper addresses the challenge of incorporating visual context into conversational AI by introducing a new visually conditioned Future Utterance Prediction task. By training a co-attentional multimodal video transformer on instructional videos, the model outperforms text-only baselines and achieves state-of-the-art performance on several VideoQA benchmarks.
While most conversational AI systems focus on textual dialogue only, conditioning utterances on visual context (when it's available) can lead to more realistic conversations. Unfortunately, a major challenge for incorporating visual context into conversational dialogue is the lack of large-scale labeled datasets. We provide a solution in the form of a new visually conditioned Future Utterance Prediction task. Our task involves predicting the next utterance in a video, using both visual frames and transcribed speech as context. By exploiting the large number of instructional videos online, we train a model to solve this task at scale, without the need for manual annotations. Leveraging recent advances in multimodal learning, our model consists of a novel co-attentional multimodal video transformer, and when trained on both textual and visual context, outperforms baselines that use textual inputs alone. Further, we demonstrate that our model trained for this task on unlabelled videos achieves state-of-the-art performance on a number of downstream VideoQA benchmarks such as MSRVTT-QA, MSVD-QA, ActivityNet-QA and How2QA.