Early detection of slight bruises in apples by cost-efficient near-infrared imaging

Near-infrared (NIR) spectroscopy has been widely reported for its useful applications in assessing internal fruit qualities. Motivated by apple consumption in the global market, this study aims to evaluate the possibility of applying NIR imaging to detect slight bruises in apple fruits. A simple optical setup was designed, and low-cost system components were used to promote the future development of practical and cost-efficient devices. To evaluate the effectiveness of the proposed approach, slight bruises were created by a mild impact with a comparably low impact energy of only 0.081 Joules. Experimental results showed that 100% of bruises in Jazz and Gala apples were accurately detected immediately after bruising and within 3 hours of storage. Thus, it is promising to develop customer devices to detect slight bruises for not only apple fruits but also other fruits with soft and thin skin at their early damage stages.

Photoluminescence Imaging for the In-Line Quality Control of Thin-Film Solar Cells

Renewable energy sources such as photovoltaic (PV) technologies are considered to be key drivers towards climate neutrality. Thin-film PVs, and particularly copper indium gallium selenide (CIGS) technologies, will play a crucial role in the turnaround in energy policy due to their high efficiencies, high product flexibility, light weight, easy installation, lower labour-intensiveness, and lower carbon footprint when compared to silicon solar cells. Nonetheless, challenges regarding the CIGS fabrication process such as moderate reproducibility and process tolerance are still hindering a broad market penetration. Therefore, cost-efficient and easily implementable in-line process control methods are demanded that allow for identification and elimination of non-conformal cells at an early production step. As part of this work, a practical approach towards industrial in-line photoluminescence (PL) imaging as a contact-free quality inspection tool is presented. Performance parameters of 10 CIGS samples with 32 individually contacted cells each were correlated with results from PL imaging using green and red excitation light sources. The data analysis was fully automated using Python-based image processing, object detection, and non-linear regression modelling. Using the red excitation light source, the presented PL imaging and data processing approach allows for a quantitative assessment of the cell performance.

Endometriosis in the Mouse: Challenges and Progress Toward a ‘Best Fit’ Murine Model

Endometriosis is a prevalent gynecologic condition associated with pelvic pain and infertility characterized by the implantation and growth of endometrial tissue displaced into the pelvis via retrograde menstruation. The mouse is a molecularly well-annotated and cost-efficient species for modeling human disease in the therapeutic discovery pipeline. However, as a non-menstrual species with a closed tubo-ovarian junction, the mouse poses inherent challenges as a preclinical model for endometriosis research. Over the past three decades, numerous murine models of endometriosis have been described with varying degrees of fidelity in recapitulating the essential pathophysiologic features of the human disease. We conducted a search of the peer-reviewed literature to identify publications describing preclinical research using a murine model of endometriosis. Each model was reviewed according to a panel of ideal model parameters founded on the current understanding of endometriosis pathophysiology. Evaluated parameters included method of transplantation, cycle phase and type of tissue transplanted, recipient immune/ovarian status, iterative schedule of transplantation, and option for longitudinal lesion assessment. Though challenges remain, more recent models have incorporated innovative technical approaches such as in vivo fluorescence imaging and novel hormonal preparations to overcome the unique challenges posed by murine anatomy and physiology. These models offer significant advantages in lesion development and readout toward a high-fidelity mouse model for translational research in endometriosis.

Optimization of capture protocols across species targeting up to 32000 genes and their extension to pooled DNA

In-solution based capture is becoming a method of choice for sequencing targeted sequence. We assessed and optimized a capture protocol in 20 different species from 6 different plant genus using kits from 20,000 to 200,000 baits targeting from 300 to 32,000 genes. We evaluated both the effectiveness of the capture protocol and the fold enrichment in targeted sequences. We proposed a protocol with multiplexing up to 96 samples in a single hybridization and showed it was an efficient and cost-effective strategy. We also extended the use of capture to pools of 100 samples and proved the efficiency of the method to assess allele frequency. Using a set of various organisms with different genome sizes, we demonstrated a correlation between the percentage of on-target reads vs. the relative size of the targeted sequences. Altogether, we proposed methods, strategies, cost-efficient protocols and statistics to better evaluate and more effectively use hybridization capture.

Landslide Trail Extraction Using Fire Extinguishing Model

Landslide trails are important elements of landslide inventory maps, providing valuable information for landslide risk and hazard assessment. Compared with traditional manual mapping, skeletonization methods offer a more cost-efficient way to map landslide trails, by automatically generating centerlines from landslide polygons. However, a challenge to existing skeletonization methods is that expert knowledge and manual intervention are required to obtain a branchless skeleton, which limits the applicability of these methods. To address this problem, a new workflow for landslide trail extraction (LTE) is proposed in this study. To avoid generating redundant branches and to improve the degree of automation, two endpoints, i.e., the crown point and the toe point, of the trail were determined first, with reference to the digital elevation model. Thus, a fire extinguishing model (FEM) is proposed to generate skeletons without redundant branches. Finally, the effectiveness of the proposed method is verified, by extracting landslide trails from landslide polygons of various shapes and sizes, in two study areas. Experimental results show that, compared with the traditional grassfire model-based skeletonization method, the proposed FEM is capable of obtaining landslide trails without spurious branches. More importantly, compared with the baseline method in our previous work, the proposed LTE workflow can avoid problems including incompleteness, low centrality, and direction errors. This method requires no parameter tuning and yields excellent performance, and is thus highly valuable for practical landslide mapping.

Cost Efficient GPU Cluster Management for Training and Inference of Deep Learning

Expanding the scale of GPU-based deep learning (DL) clusters would bring not only accelerated AI services but also significant energy consumption costs. In this paper, we propose a cost efficient deep learning job allocation (CE-DLA) approach minimizing the energy consumption cost for the DL cluster operation while guaranteeing the performance requirements of user requests. To do this, we first categorize the DL jobs into two classes: training jobs and inference jobs. Through the architecture-agnostic modeling, our CE-DLA approach is able to conduct the delicate mapping of heterogeneous DL jobs to GPU computing nodes. Second, we design the electricity price-aware DL job allocation so as to minimize the energy consumption cost of the cluster. We show that our approach efficiently avoids the peak-rate time slots of the GPU computing nodes by using the sophisticated mixed-integer nonlinear problem (MINLP) formulation. We additionally integrate the dynamic right-sizing (DRS) method with our CE-DLA approach, so as to minimize the energy consumption of idle nodes having no running job. In order to investigate the realistic behavior of our approach, we measure the actual output from the NVIDIA-based GPU devices with well-known deep neural network (DNN) models. Given the real trace data of the electricity price, we show that the CE-DLA approach outperforms the competitors in views of both the energy consumption cost and the performance for DL job processing.

Sampling Designs for Monitoring Ichthyoplankton in the Estuary Area: A Case Study on Coilia mystus in the Yangtze Estuary

A fishery-independent survey can provide detailed information for fishery assessment and management. However, the sampling design for the survey on ichthyoplankton in the estuary area is still poorly understood. In this study, we developed six stratified schemes with various sample sizes, attempting to find cost-efficient sampling designs for monitoring Coilia mystus ichthyoplankton in the Yangtze Estuary. The generalized additive model (GAM) with the Tweedie distribution was used to quantify the “true” distribution of C. mystus eggs and larvae, based on the data from the fishery-independent survey in 2019–2020. The performances of different sampling designs were evaluated by relative estimation error (REE), relative bias (RB), and coefficient of variation (CV). The results indicated that appropriate stratifications with intra-stratum homogeneity and inter-stratum heterogeneity could improve precision. The stratified schemes should be divided not only between the North Branch and South Branch but between river and sea. No less than two stratifications in the South Branch could also get better performance. The sample sizes of 45–55 were considered as the cost-efficient range. Compared to other monitoring programs, monitoring ichthyoplankton in the estuary area required a more complex stratification and a higher resolution sampling. The design ideology and optimization methodology in our study would provide references to sampling designs for ichthyoplankton in the estuary area.