Enhancing automated strabismus classification with limited data: Data augmentation using StyleGAN2-ADA.

In this study, we propose a generative data augmentation technique to overcome the challenges of severely limited data when designing a deep learning-based automated strabismus diagnosis system. We implement a generative model based on the StyleGAN2-ADA model for system design and assess strabismus classification performance using two classifiers. We evaluate the capability of our proposed method against traditional data augmentation techniques and confirm a substantial enhancement in performance. Furthermore, we conduct experiments to explore the relationship between the diagnosis agreement among ophthalmologists and the generation performance of the generative model. Beyond FID, we validate the generative samples on the classifier to establish their practicality. Through these experiments, we demonstrate that the generative model-based data augmentation improves overall quantitative performance in scenarios of extreme data scarcity and effectively mitigates overfitting issues during deep learning model training.

An efficient tomato-detection method based on improved YOLOv4-tiny model in complex environment.

Automatic and accurate detection of fruit in greenhouse is challenging due to complicated environment conditions. Leaves or branches occlusion, illumination variation, overlap and cluster between fruits make the fruit detection accuracy to decrease. To address this issue, an accurate and robust fruit-detection algorithm was proposed for tomato detection based on an improved YOLOv4-tiny model. First, an improved backbone network was used to enhance feature extraction and reduce overall computational complexity. To obtain the improved backbone network, the BottleneckCSP modules of the original YOLOv4-tiny backbone were replaced by a Bottleneck module and a reduced version of BottleneckCSP module. Then, a tiny version of CSP-Spatial Pyramid Pooling (CSP-SPP) was attached to the new backbone network to improve the receptive field. Finally, a Content Aware Reassembly of Features (CARAFE) module was used in the neck instead of the traditional up-sampling operator to obtain a better feature map with high resolution. These modifications improved the original YOLOv4-tiny and helped the new model to be more efficient and accurate. The experimental results showed that the precision, recall, F 1 score, and the mean average precision (mAP) with Intersection over Union (IoU) of 0.5 to 0.95 were 96.3%, 95%, 95.6%, and 82.8% for the improved YOLOv4-tiny model, respectively. The detection time was 1.9 ms per image. The overall detection performance of the improved YOLOv4-tiny was better than that of state-of-the-art detection methods and met the requirements of tomato detection in real time.

An Ultrasonic Object Detection Applying the ID Based on Spread Spectrum Technique for a Vehicle.

When an ultrasonic sensor generates an ultrasonic wave and detects an obstacle from a reflected wave, a signal transmitted by other ultrasonic sensors would be interference. In this paper, to overcome the interference, a transducer transmits a signal with a unique ID modulated. The interference is ignored by verifying that the reflected signal includes its ID. The ID verification process uses a correlation between the received signal and the ID. Therefore, the ID is selected from orthogonal codes with good cross-correlation. Long code has the advantage of being more robust to interference. However, the reflected wave from nearby obstacles might return before the transmission ends. Therefore, the 7-bit Barker code is applied for near obstacle detection and a 31-bit Gold code is used for distant obstacle detection. The modulation technique is DQPSK, which is available in a narrow bandwidth and has a simple receiver structure. In ID recognition based on correlation, a near-far problem occurs due to a large amplitude difference between the received wave and interference. The addition of a zero-crossing detector solves this problem. The hardware is implemented based on the algorithm proposed in this paper. The simulation showed a detection rate of at least 90% and the the result of the real measurement represented a detection rate of 97.3% at 0.5 m and 94.5% at 2 m.

Hybrid precoding based on matrix-adaptive method for multiuser large-scale antenna arrays.

Massive multiple-input multiple-output (MIMO) is envisioned to offer a considerable improvement in capacity, but it has a high cost and the radio frequency (RF) chain components have a high power consumption at high frequency. To address this problem, a hybrid analog and digital precoding scheme has been studied recently, which restricts the number of RF chains to far less than the number of antenna elements. In this paper, we consider the downlink communication of a massive multiuser multiple-input single-output (MU-MISO) system and propose an iterative hybrid precoding algorithm to approach the capacity performance of the traditional full digital precoding scheme. We aim to attain a large baseband gain by zero-forcing (ZF) digital precoding on the equivalent channel and then minimize the total power to obtain the optimal RF precoder. Simulation results show that the proposed method can approach the performance of the conventional fully digital precoding with a low computational complexity.