Bolin Zhang (张博麟)

This paper explores the task of video moment retrieval (VMR), which aims to localize the temporal boundary of a specific moment from an untrimmed video by a sentence query. Previous methods either extract pre-defined candidate moment features and select the moment that best matches the query by ranking, or directly align the boundary clips of a target moment with the query and predict matching scores. Despite their effectiveness, these methods mostly focus only on aligning the query and single-level clip or moment features, and ignore the different granularities involved in the video itself, such as clip, moment, or video, resulting in insufficient cross-modal interaction. To this end, we propose a Temporal Localization Network with Hierarchical Contrastive Learning (HCLNet) for the VMR task. Specifically, we introduce a hierarchical contrastive learning method to better align the query and video by maximizing the mutual information (MI) between query and three different granularities of video to learn informative representations. Meanwhile, we introduce a self-supervised cycle-consistency loss to enforce the further semantic alignment between fine-grained video clips and query words. Experiments on three standard benchmarks show the effectiveness of our proposed method.

According to the given natural language query, moment retrieval aims to localize the most relevant moment in an untrimmed video. The existing solutions for this problem can be roughly divided into two categories based on whether candidate moments are generated: i) Moment-based approach: It pre-cuts the video into a set of candidate moments, performs multimodal fusion, and evaluates matching scores with the query. ii) Clip-based approach: It directly aligns video clips and query with predicting matching scores without generating candidate moments. Both frameworks have respective shortcomings: the moment-based models suffer from heavy computations, while the performance of clip-based models is familiarly inferior to moment-based counterparts. To this end, we design an intuitive and efficient Dual-Channel Localization Network (DCLN) to balance computational cost and retrieval performance. For reducing computational cost, we capture the temporal relations of only a few video moments with the same start or end boundary in the proposed dual-channel structure. The start or end channel map index represents the corresponding video moment’s start or end time boundary. For improving model performance, we apply the proposed dual-channel localization network to efficiently encode the temporal relations on the dual-channel map and learn discriminative features to distinguish the matching degree between natural language query and video moments. The extensive experiments on two standard benchmarks demonstrate the effectiveness of our proposed method.

Cross-modal recipe retrieval is vital for transforming visual food cues into actionable cooking guidance, making culinary creativity more accessible. Existing methods separately encode the recipe Title, Ingredient, and Instruction using different text encoders, then aggregate them to obtain recipe feature, and finally match it with encoded image feature in a joint embedding space. These methods perform well but require significant computational cost. In addition, they only consider matching the entire recipe and the image but ignore the fine-grained correspondence between recipe components and the image, resulting in insufficient cross-modal interaction. To this end, we propose Unified Text Encoder with Fine-grained Contrastive Learning (UTE-FCL) to achieve a simple but efficient model. Specifically, in each recipe, UTE-FCL first concatenates each of the Ingredient and Instruction texts composed of multiple sentences as a single text. Then, it connects these two concatenated texts with the original single-phrase Title to obtain the concatenated recipe. Finally, it encodes these three concatenated texts and the original Title by a Transformer-based Unified Text Encoder (UTE). This proposed structure greatly reduces the memory usage and improves the feature encoding efficiency. Further, we propose fine-grained contrastive learning objectives to capture the correspondence between recipe components and the image at Title, Ingredient, and Instruction levels by measuring the mutual information. Extensive experiments demonstrate the effectiveness of UTE-FCL compared to existing methods.