Zero-shot Human-Object Interaction (HOI) detection aims to identify both seen and unseen HOI categories in an image. Most existing methods rely on semantic knowledge distilled from CLIP to find novel interactions but fail to fully exploit the powerful generalization ability of vision-language models, leading to impaired transferability. In this paper, we introduce a novel framework for zero-shot HOI detection. We first utilize vision-language models (VLMs) to generate visual captions from multiple perspectives, including humans, objects, and environments, to enhance interaction understanding. Then, we propose a multi-modal fusion encoder to fully leverage these visual captions. Additionally, to equip the HOI detector with a thorough consideration of contextual information in the image, we design a novel multi-branch HOI network that aggregates features at the instance, union, and global levels. Experiments on prevalent benchmarks demonstrate that our model achieves promising performance under a variety of zero-shot settings. The source codes are available at https://github.com/aqingcv/VC-HOI.

Disassembly production lines that employ shared stations with multiple workers and robots are ideal for addressing obsolete products with complex structures and hazardous parts. In this study, a multiproduct multi-man–robot disassembly line balancing problem (MPMMR-DLBP) is developed and its mixed-integer programming model (MIPM) is established to minimize the number of stations, idle balancing index of operators (workers and robots), and the number of operators. In addition, a problem-oriented group evolutionary (POGE) algorithm is proposed to efficiently solve the MPMMR-DLBP. The proposed POGE develops a new “1+3” encoding mode and a heuristic decoding strategy based on the shortest time to complete tasks to construct a one-to-one correspondence between encoding sequences and disassembly schemes. Moreover, a reassociation evolution operation (REO) and a mapping crossover operation (MCO) are designed to generate new solutions and allow the population to evolve to the global optimum. Subsequently, the correctness of MIPM and the performance of POGE are verified using two small-scale cases. Finally, an actual MPMMR-DLBP for the mixed disassembly of refrigerators, microwave ovens, and dishwashers is optimized by POGE. A comparison of the optimized results with the other three common algorithms shows that the POGE is superior in the large-scale MPMMR-DLBP, and multiple optimized disassembly schemes are provided for decision makers.

To reasonably arrange disassembly facilities and plan enterprise space, we propose a parallel disassembly line balancing problem (PW-PDLBP) with parallel workstations. Additionally, a mixed-integer nonlinear programming (MINLP) model that minimizes the line length, number of workstations, idle time balancing index, and energy consumption is established based on the problem characteristics and is solved using the GUROBI optimizer. Furthermore, a multiobjective enhanced differential evolution algorithm (MEDE) is developed to obtain high-quality disassembly schemes for PW-PDLBP. The correctness of encoding and decoding and the solving performance of MEDE are verified by comparing with the MINLP model and four existing algorithms. Then, an instance consisting of two different types of end-of-life TVs is optimized. Finally, the effectiveness of PW-PDLBP in improving enterprise space utilization is validated by comparing it with the parallel line layout without parallel workstations.