New technologies including big data and intelligence emergence etc., have a significant impact on education and teaching. Physics, as the foundation of natural science, plays a vital role in promoting the origin and development of artificial intelligence via its discipline ideas and methods. In order to follow the development trends of frontier technology and the pulse of The Times, in this paper, the reform attempts in the teaching concept, teaching content and teaching mode of college physics, have been discussed based on the course idea of “AI+”. And we propose some innovation measures including changing teaching concept to embrace artificial intelligence, using AI assist teaching to improve AI literacy and construct curriculum knowledge maps, and deepening the connotation and extension of teaching content. That is so as to improve the quality of college physics and others teaching in the new era.

A U-shaped network is suggested for anomaly behavior recognition based on Mamba in order to overcome difficulties in unsupervised anomaly behavior detection, such as the predictor’s propensity for abnormal generalization and the target objects' scale disparities. The network improves both global and local features to constrain undesired generalization ability in predictions. A state space model is introduced in the encoder to strengthen the extraction of global features. A multi-scale spatial channel fusion (M-SCF) strategy is designed to integrate feature information from different receptive fields, thereby reducing the interference of scale differences on local features. Skip connections are used in the decoder to enrich shallow feature information and enhance the ability to capture contextual information. The proposed method has been extensively validated on the UCSD Ped2, Avenue, and Shanghai Tech datasets, with respective recognition accuracies of 98.1%, 89.8%, and 78.5%. Mamba can successfully increase the accuracy of abnormal behavior identification, as evidenced by the findings, which demonstrate superior accuracy when compared to several sophisticated algorithms in recent years.

In order to overcome the difficulty of crowd counting in low light, RGB-T crowd counting attempts to create maps of crowd density utilizing complimentary information from visual and thermal imagery. However, existing RGB-T crowd counting methods face issues such as scale variation and background interference during cross-modality information fusion. To tackle these challenges, we propose an RGB-T crowd counting method based on multi-scale perception and infrared feature enhancement (MSENet). Our approach presents an RGB-T feature fusion mechanism (RTFM) that creates an infrared enhancement structure to completely capture crowd information in thermal images and uses a multi-branch structure for multi-scale feature extraction. Additionally, we utilize dense connections and information divergence mechanisms to transfer complementary features to each modality, achieving a reusable expression of complementary features and enhanced modality features. We evaluate our proposed method on the RGBT-CC dataset and the ShanghaiTechRGBD dataset through comparative experiments. The results demonstrate that our method outperforms existing state-of-the-art approaches on the RGBT-CC dataset, exhibiting good accuracy, robustness and good generalization.

To solve problems including low-quality audio and video modal features and inadequate interaction between various modalities, a multimodal sentiment analysis approach based on cross-modal Transformer (CMT) and audio and video feature optimization is suggested. Firstly, we propose a audio and video features optimizing mechanism (AVFOM), which increases the density of sentiment information in audio and video features through synergistic interaction with textual features, thereby improving the quality of audio and video features. Secondly, in order to accomplish full interaction between text-audio and text-video modalities and learn consistent knowledge across various modalities, we construct a cross-modal Transformer structure with text as the dominant modality. Additionally, a label generation method based on the self-supervised learning strategy is introduced to perform single-modality sentiment prediction tasks, learning the characteristics of each modality separately. The proposed method is extensively validated and tested on two public datasets, CMU-MOSI and CMU-MOSEI, which surpass many currently advanced methods in terms of performance and effectively improve the accuracy of multimodal sentiment analysis.

This research proposes a micro expression recognition network that incorporates LBP and parallel attention method to address the issues of small feature discrimination, background noise interference, and weak intensity of facial micro-expression changes. The network inputs the RGB image into the densely connected improved Shuffle Stage branch to extract the global features of the face and enhance the association of contextual semantic information. The LBP image is input into the local texture feature branch composed of a multi-scale layered convolutional neural network to extract detailed information. Following extraction of the dual-branch feature, the network backend implements a parallel attention technique to enhance feature fusion capabilities, reduce background noise, and concentrate on the micro-expression feature's interest region. The proposed method is tested on three public data sets including CASME, CASME II and SMIC, and the recognition is accurate The rates reached 85.18%, 74.53% and 81.19% respectively. The experimental results show that the proposed method effectively improves the accuracy of micro expression recognition, which is better than many current advanced methods.