How to work with Cross-Attention part2(Machine Learning)
- Multispectral Pedestrian Detection via Reference Box Constrained Cross Attention and Modality Balanced Optimization(arXiv)
Author : Yinghui Xing, Song Wang, Guoqiang Liang, Qingyi Li, Xiuwei Zhang, Shizhou Zhang, Yanning Zhang
Abstract : Multispectral pedestrian detection is an important task for many around-the-clock applications, since the visible and thermal modalities can provide complementary information especially under low light conditions. To reduce the influence of hand-designed components in available multispectral pedestrian detectors, we propose a MultiSpectral pedestrian DEtection TRansformer (MS-DETR), which extends deformable DETR to multi-modal paradigm. In order to facilitate the multi-modal learning process, a Reference box Constrained Cross-Attention (RCCA) module is firstly introduced to the multi-modal Transformer decoder, which takes fusion branch together with the reference boxes as intermediaries to enable the interaction of visible and thermal modalities. To further balance the contribution of different modalities, we design a modality-balanced optimization strategy, which aligns the slots of decoders by adaptively adjusting the instance-level weight of three branches. Our end-to-end MS-DETR shows superior performance on the challenging KAIST and CVC-14 benchmark dataset
2. Cross-modal Attention Congruence Regularization for Vision-Language Relation Alignment(arXiv)
Author : Rohan Pandey, Rulin Shao, Paul Pu Liang, Ruslan Salakhutdinov, Louis-Philippe Morency
Abstract : Despite recent progress towards scaling up multimodal vision-language models, these models are still known to struggle on compositional generalization benchmarks such as Winoground. We find that a critical component lacking from current vision-language models is relation-level alignment: the ability to match directional semantic relations in text (e.g., “mug in grass”) with spatial relationships in the image (e.g., the position of the mug relative to the grass). To tackle this problem, we show that relation alignment can be enforced by encouraging the directed language attention from ‘mug’ to ‘grass’ (capturing the semantic relation ‘in’) to match the directed visual attention from the mug to the grass. Tokens and their corresponding objects are softly identified using the cross-modal attention. We prove that this notion of soft relation alignment is equivalent to enforcing congruence between vision and language attention matrices under a ‘change of basis’ provided by the cross-modal attention matrix. Intuitively, our approach projects visual attention into the language attention space to calculate its divergence from the actual language attention, and vice versa. We apply our Cross-modal Attention Congruence Regularization (CACR) loss to UNITER and improve on the state-of-the-art approach to Winoground.
