Fish feeding behavior analysis plays a critical role in aquaculture, marine fisheries, and ecological monitoring. At present, the accuracy of current fish feeding behavior analysis is easily affected by the environment, and the hypoxia state in water and high temperature have a huge impact on behavior. In addition, the poor stability, and the speed is affected by the calculation amount of the model, so it cannot be applied to aquaculture ponds. In order to solve the above problems, we propose an improved RepVGG-based method for fish feeding behavior analysis to identify nine behavioral states of fish populations, including hypoxia, hot, and normal states of none, medium, and strong feeding. Firstly, our method adds the identity and residual branches to the block of the VGG network, focusing on the acceleration operation to ensure accuracy. At the same time, the Efficient Channel Attention (ECA) module is added to reduce dimensionality to balance speed and accuracy. To evaluate the effectiveness of the method, it is validated on a constructed fish feeding behavior dataset and compared with a Convolutional Neural Network (CNN) including AlexNet, VGG16, ResNet34, MobileNet V3, and RepVGG. The experimental results show that after 15 epochs of training, the average precision of proposed method reached 97%. Compared with the basic classification algorithm, our method can increase by at least 3% while the inference speed exceeds 85 Frames Per Second (FPS). This study can be integrated into the aquaculture vision system to guide users to plan feeding strategies, and provide new ideas and methods for water quality monitoring.

Multimodal Prompt Learning (MPL) has emerged as a pivotal technique for adapting large-scale Visual Language Models (VLMs). However, current MPL methods are fundamentally limited by their optimization of a single, static point representation. This paradigm is inherently brittle, leads to overfitting on base classes, and generalizes poorly to novel or ambiguous categories. We challenge this point paradigm, proposing that robust generalization requires learning a semantic cloud (i.e., a distribution over the embedding space). To achieve this, we introduce Points-to-Clouds (P2C), a novel framework inspired by diffusion models that reframes prompt learning as a dynamic denoising task. At the core of P2C is a dual denoising mechanism: a Dynamic Prompt Denoising (DPD) mechanism perturbs text prompts with sophisticated, annealed noise to learn a smoother semantic landscape, while an auxiliary V-L Mapper denoising loss re-tasks the mapper as a denoising autoencoder. This forces the mapper to reconstruct clean visual prompts from noisy text inputs, ensuring robust cross-modal alignment. Extensive experiments across 11 datasets demonstrate that P2C consistently outperforms strong baselines. On the base-to-novel generalization benchmark, our method achieves a Harmonic Mean of 79.7%, representing a relative improvement of 1.4% over the baseline. The code and models are available at https://vranlee.github.io/P2C/.

Multi-object tracking (MOT) is a fundamental task in computer vision with critical applications in autonomous driving and robotics. Multimodal MOT that integrates visible light and thermal infrared information is particularly essential for robust autonomous driving systems. However, effectively fusing these heterogeneous modalities is challenging. Simple strategies like concatenation or addition often fail to bridge the significant non-linear distribution gap between their feature representations, which can lead to modality conflicts and degrade tracking accuracy. Drawing inspiration from the connection between multimodal MOT and the iterative refinement in diffusion models, this paper proposes DM$^3$T, a novel framework that reformulates multimodal fusion as an iterative feature alignment process to generate accurate and temporally coherent object trajectories. Our approach performs iterative cross-modal harmonization through a proposed Cross-Modal Diffusion Fusion (C-MDF) module. In this process, features from both modalities provide mutual guidance, iteratively projecting them onto a shared, consistent feature manifold. This enables the learning of complementary information and achieves deeper fusion compared to conventional methods. Additionally, we introduce a plug-and-play Diffusion Refiner (DR) to enhance and refine the unified feature representation. To further improve tracking robustness, we design a Hierarchical Tracker that adaptively handles confidence estimation. DM$^3$T unifies object detection, state estimation, and data association into a comprehensive online tracking framework without complex post-processing. Extensive experiments on the VT-MOT benchmark demonstrate that our method achieves 41.7 HOTA, representing a 1.54% relative improvement over existing state-of-the-art methods. The code and models are available at https://vranlee.github.io/DM-3-T/.