Traditional object detection methods always assume both of the training and test data follow the same distribution, but this cannot always be guaranteed in the real world. Domain adaptive methods are proposed to handle this situation. However, existing methods generally ignore the semantic alignment at feature level when they try to align data distributions between source and target domains. In this paper, we propose a novel unsupervised cross-domain object detection method, named Cycle-consistent domain Adaptive Faster RCNN (CA-FRCNN). A couple of Generative Adversarial Nets (GAN) are used to make the features from two domains consistent at both data distribution level and semantic level. Specifically, features from source domain are transformed to the target domain. Then they are aligned with features from target domain. At the same time, target features are handled with similar operations. Furthermore, a cycle-consistent loss is optimized to guarantee that the semantic information is preserved before and after the style translations. In the end, identity module is used to make the feature in source domain equivalent to the reconstructed feature output by the source generator, whose input is from the source domain. For the feature in the target domain, similar identity is required. Experiments on multiple datasets show that our method performs better than previous state-of-the-art methods.

Background: For rehabilitation training systems, it is essential to automatically record and recognize exercises, especially when more than one type of exercise is performed without a predefined sequence. Most motion recognition methods are based on feature engineering and machine learning algorithms. Time-domain and frequency-domain features are extracted from original time series data collected by sensor nodes. For high-dimensional data, feature selection plays an important role in improving the performance of motion recognition. Existing feature selection methods can be categorized into filter and wrapper methods. Wrapper methods usually achieve better performance than filter methods; however, in most cases, they are computationally intensive, and the feature subset obtained is usually optimized only for the specific learning algorithm. Objective: This study aimed to provide a feature selection method for motion recognition of upper-limb exercises and improve the recognition performance. Methods: Motion data from 5 types of upper-limb exercises performed by 21 participants were collected by a customized inertial measurement unit (IMU) node. A total of 60 time-domain and frequency-domain features were extracted from the original sensor data. A hybrid feature selection method by combining filter and wrapper methods (FESCOM) was proposed to eliminate irrelevant features for motion recognition of upper-limb exercises. In the filter stage, candidate features were first selected from the original feature set according to the significance for motion recognition. In the wrapper stage, k-nearest neighbors (kNN), Naïve Bayes (NB), and random forest (RF) were evaluated as the wrapping components to further refine the features from the candidate feature set. The performance of the proposed FESCOM method was verified using experiments on motion recognition of upper-limb exercises and compared with the traditional wrapper method. Results: Using kNN, NB, and RF as the wrapping components, the classification error rates of the proposed FESCOM method were 1.7%, 8.9%, and 7.4%, respectively, and the feature selection time in each iteration was 13 seconds, 71 seconds, and 541 seconds, respectively. Conclusions: The experimental results demonstrated that, in the case of 5 motion types performed by 21 healthy participants, the proposed FESCOM method using kNN and NB as the wrapping components achieved better recognition performance than the traditional wrapper method. The FESCOM method dramatically reduces the search time in the feature selection process. The results also demonstrated that the optimal number of features depends on the classifier. This approach serves to improve feature selection and classification algorithm selection for upper-limb motion recognition based on wearable sensor data, which can be extended to motion recognition of more motion types and participants.

The approach of image processing and deep learning has shown to be a breakthrough in the field of medical image diagnosis such as dermoscopic image analysis for skin cancer recognition and their classification. Skin cancer cases are increasing every year and pose a significant threat for health. In recent studies, convolutional neural network (CNN) has accomplished remarkable success in classifying skin cancer images. CNN is limited to extracting features from minor objects from input dermoscopic image and fails to pinpoint significant regions. Consequently, the researchers of this study have utilized vision transformers (VIT), known for their robust performance in conventional classification assignments. The self-attention mechanism (SAM) aims to enhance the significance of pivotal characteristics while modifying the influence of noise-inducing features. Specifically, an enhanced transformer network architecture has been introduced in this context. To assess its effectiveness, several enhancements have been applied to the model. Initially, a ViT network is implemented to evaluate its efficacy in identifying skin cancer. Subsequently, Adaptive-scale image fragmentation is utilized to sequentially process the image, emphasizing adaptive-scale features through patch embedding. Furthermore, contrastive learning is employed to ensure that similar skin cancer data is encoded differently, aiming for distinct encoding outcomes for different data. Skin cancer dataset namely ISIC 2019 is retrieved in this study, locally accessible at Kaggle’s official website. This dataset consists of dermoscopic images of skin cancer having several types: dermatofibroma, melanoma, actinic keratosis, basal cell carcinoma, nevus, vascular lesion, and pigmented benign keratosis. The ViT model has achieved 99.66% accuracy, 94.85% precision, 93.74% recall, and 94.52% f1-score. Three deep learning models Inception V3, MobileNet, and ResNet-50 were also applied with transfer learning approach as comparison to proposed ViT model for performance evaluation that resulted in 72%, 94.3, and 89% accuracies, respectively. The transformer network has shown remarkable success in natural language processing and in the domain of image analysis. These achievements establish a solid groundwork to classify skin cancer using multimodal data. This paper is confident to captivate the attention of medical researchers, computer engineers, dermatologists, and scholars across various related disciplines. Its insights promise to offer enhanced convenience for patients in their respective fields.