ABSTRACT
This paper delves into image detection based on distributed deep-learning techniques for intelligent traffic systems or self-driving cars. The accuracy and precision of neural networks deployed on edge devices (e.g., CCTV (closed-circuit television) for road surveillance) with small datasets may be compromised, leading to the misjudgment of targets. To address this challenge, TensorFlow and PyTorch were used to initialize various distributed model parallel and data parallel techniques. Despite the success of these techniques, communication constraints were observed along with certain speed issues. As a result, a hybrid pipeline was proposed, combining both dataset and model distribution through an all-reduced algorithm and NVlinks to prevent miscommunication among gradients. The proposed approach was tested on both an edge cluster and Google cluster environment, demonstrating superior performance compared to other test settings, with the quality of the bounding box detection system meeting expectations with increased reliability. Performance metrics, including total training time, images/second, cross-entropy loss, and total loss against the number of the epoch, were evaluated, revealing a robust competition between TensorFlow and PyTorch. The PyTorch environment's hybrid pipeline outperformed other test settings.
ABSTRACT
Tomato cultivation in the greenhouse can be facilitated by supplemental light. We compared the combined effects of nutrients, water, and supplemental light (red) on tomato fruit quality. To do this, three different nutrient conditions were tested, i.e., (1) low N, (2) standard N, and (3) high N. Water was supplied either at -30 kPa (sufficient) or -80 kPa (limited) of soil water potential. Supplemental red LED light was turned either on or off. The metabolites from tomato fruits were profiled using non-targeted mass spectrometry (MS)-based metabolomic approaches. The lycopene content was highest in the condition of high N and limited water in the absence of supplemental light. In the absence of red lighting, the lycopene contents were greatly affected by nutrient and water conditions. Under the red lighting, the nutrient and water conditions did not play an important role in enhancing lycopene content. Lower N resulted in low amino acids. Low N was also likely to enhance some soluble carbohydrates. Interestingly, the combination of low N and red light led to a significant increase in sucrose, maltose, and flavonoids. In high N soil, red light increased a majority of amino acids, including aspartic acid and GABA, and sugars. However, it decreased most of the secondary metabolites such as phenylpropanoids, polyamines, and alkaloids. The water supply effect was minor. We demonstrated that different nutrient conditions of soil resulted in a difference in metabolic composition in tomato fruits and the effect of red light was variable depending on nutrient conditions.
