The definition of vehicle viewpoint annotations is ambiguous due to human subjective judgment, which makes the cross-domain vehicle re-identification methods unable to learn the viewpoint invariance features during source domain pre-training. This will further lead to cross-view misalignment in downstream target domain tasks. To solve the above challenges, this paper presents a dual-level viewpoint-learning framework that contains an angle invariance pre-training method and a meta-orientation adaptation learning strategy. The dual-level viewpoint-annotation proposal is first designed to concretely redefine the vehicle viewpoint from two aspects (i.e., angle-level and orientation-level). An angle invariance pre-training method is then proposed to preserve identity similarity and difference across the cross-view; this consists of a part-level pyramidal network and an angle bias metric loss. Under the supervision of angle bias metric loss, the part-level pyramidal network, as the backbone, learns the subtle differences of vehicles from different angle-level viewpoints. Finally, a meta-orientation adaptation learning strategy is designed to extend the generalization ability of the re-identification model to the unseen orientation-level viewpoints. Simultaneously, the proposed meta-learning strategy enforces meta-orientation training and meta-orientation testing according to the orientation-level viewpoints in the target domain. Extensive experiments on public vehicle re-identification datasets demonstrate that the proposed method combines the redefined dual-level viewpoint-information and significantly outperforms other state-of-the-art methods in alleviating viewpoint variations.

With the emergence of vision-language pre-trained models, such as CLIP, some textual prompts have been gradually introduced recently into re-identification (Re-ID) tasks to obtain considerably robust multimodal information. However, most textual descriptions based on vehicle Re-ID tasks only contain identity index words without specific words to describe vehicle view information, thereby resulting in difficulty to be widely applied in vehicle Re-ID tasks with view variations. This case inspires us to propose a CLIP-driven view-aware prompt learning framework for unsupervised vehicle Re-ID. We first design a learnable textual prompt template called view-aware context optimization (ViewCoOp) based on dynamic multi-view word embeddings, which can fully obtain the proportion and position encoding of each view in the whole vehicle body region. Subsequently, a cross-modal mutual graph is constructed to explore the connections between inter-modal and intra-modal. Each sample is treated as a graph node, which extracts textual features based on ViewCoOp and the visual features of images. Moreover, leveraging the inter-cluster and intra-cluster correlation in the bimodal clustering results in the determination of connectivity between graph node pairs. Lastly, the proposed cross-modal mutual graph method utilizes supervised information from the bimodal gap to directly fine-tune the image encoder of CLIP for downstream unsupervised vehicle Re-ID tasks. Extensive experiments verify that the proposed method is capable of effectively obtaining cross-modal description ability from multiple views.