Infrared-Visible Cross-Modal Person Re-Identification with an X Modality

This paper focuses on the emerging Infrared-Visible cross-modal person re-identification task (IV-ReID), which takes infrared images as input and matches with visible color images. IV-ReID is important yet challenging, as there is a significant gap between the visible and infrared images. To reduce this ‘gap’, we introduce an auxiliary X modality as an assistant and reformulate infrared-visible dual-mode cross-modal learning as an X-Infrared-Visible three-mode learning problem. The X modality restates from RGB channels to a format with which cross-modal learning can be easily performed. With this idea, we propose an X-Infrared-Visible (XIV) ReID cross-modal learning framework. Firstly, the X modality is generated by a lightweight network, which is learnt in a self-supervised manner with the labels inherited from visible images. Secondly, under the XIV framework, cross-modal learning is guided by a carefully designed modality gap constraint, with information exchanged cross the visible, X, and infrared modalities. Extensive experiments are performed on two challenging datasets SYSU-MM01 and RegDB to evaluate the proposed XIV-ReID approach. Experimental results show that our method considerably achieves an absolute gain of over 7% in terms of rank 1 and mAP even compared with the latest state-of-the-art methods.

Form deviation and constraints on productivity

Issues concerning the nature of so-called ‘filler-gap’ constructions and the constraints on their acceptability in different languages have for a long time been at the center of the linguistic debate. In more recent years, claims have often been made that many or all constraints on such constructions can be explained in terms of processing restrictions, properties of information structure, or limits to attention. However, in the present paper I will argue that at least some of these constraints cannot be explained by such factors. Instead, I will claim that speakers’ judgements of acceptability could be negatively affected if the syntactic structure of a productively formed instance of a schematic construction is not sufficiently similar to the prototypical entrenched structure of the same construction. Importantly, this might happen whether the relevant form is a phonological pattern, or a pattern of schematic syntactic categories. I will show how the proposed principles have the potential to explain a number of well-known constraints which have generally not been regarded as related. More specifically, I will show how the principles can account for some striking similarities between filler-gap constraints in English and Swedish, as well as some unexpected differences. I will also argue that the filler-gap constraint referred to as the complementizer-gap effect can be explained by the same principles that are involved in phenomena such as restrictions on wanna-contraction in English and adjective defectiveness in Swedish. The analysis will be based on the constructional approaches of Croft (2001) and Verhagen (2009), and some crucial issues will concern the relationship between syntactic and phonological structure in a construction, and the notions of non-iconic constituency and gradient syntactic attachment.