Bottom-Up and Top-Down Attention for Image Captioning and Visual Question Answering
This work addresses the problem of fine-grained visual understanding for image captioning and VQA systems, representing a strong incremental improvement over existing attention methods.
The authors tackled image understanding tasks by proposing a combined bottom-up and top-down attention mechanism that operates at the level of objects and salient regions, achieving a new state-of-the-art on MSCOCO image captioning with CIDEr/SPICE/BLEU-4 scores of 117.9/21.5/36.9 and first place in the 2017 VQA Challenge.
Top-down visual attention mechanisms have been used extensively in image captioning and visual question answering (VQA) to enable deeper image understanding through fine-grained analysis and even multiple steps of reasoning. In this work, we propose a combined bottom-up and top-down attention mechanism that enables attention to be calculated at the level of objects and other salient image regions. This is the natural basis for attention to be considered. Within our approach, the bottom-up mechanism (based on Faster R-CNN) proposes image regions, each with an associated feature vector, while the top-down mechanism determines feature weightings. Applying this approach to image captioning, our results on the MSCOCO test server establish a new state-of-the-art for the task, achieving CIDEr / SPICE / BLEU-4 scores of 117.9, 21.5 and 36.9, respectively. Demonstrating the broad applicability of the method, applying the same approach to VQA we obtain first place in the 2017 VQA Challenge.