More Like this

1. Reduced Graphene Oxide-Metalloporphyrin Sensors for Human Breath Screening

   https://doi.org/10.3390/app112311290  

   Lee, Bo Mi ; Eetemadi, Ameen ; Tagkopoulos, Ilias ( December 2021 , Applied Sciences)

   The objective of this study is to validate reduced graphene oxide (RGO)-based volatile organic compounds (VOC) sensors, assembled by simple and low-cost manufacturing, for the detection of disease-related VOCs in human breath using machine learning (ML) algorithms. RGO films were functionalized by four different metalloporphryins to assemble cross-sensitive chemiresistive sensors with different sensing properties. This work demonstrated how different ML algorithms affect the discrimination capabilities of RGO–based VOC sensors. In addition, an ML-based disease classifier was derived to discriminate healthy vs. unhealthy individuals based on breath sample data. The results show that our ML models could predict the presence of disease-related VOC compounds of interest with a minimum accuracy and F1-score of 91.7% and 83.3%, respectively, and discriminate chronic kidney disease breath with a high accuracy, 91.7%. 

   more » « less
2. Contextual Attention for Hand Detection in the Wild

   Narasimhaswamy, S. ( October 2019 , IEEE International Conference on Computer Vision workshops)

   We present Hand-CNN, a novel convolutional network architecture for detecting hand masks and predicting hand orientations in unconstrained images. Hand-CNN extends MaskRCNN with a novel attention mechanism to incorporate contextual cues in the detection process. This attention mechanism can be implemented as an efficient network module that captures non-local dependencies between features. This network module can be inserted at different stages of an object detection network, and the entire detector can be trained end-to-end. We also introduce a large-scale annotated hand dataset containing hands in unconstrained images for training and evaluation. We show that Hand-CNN outperforms existing methods on several datasets, including our hand detection benchmark and the publicly available PASCAL VOC human layout challenge. We also conduct ablation studies on hand detection to show the effectiveness of the proposed contextual attention module. 

   more » « less
3. YOLO-LITE: A Real-Time Object Detection Algorithm Optimized for Non-GPU Computers

   https://doi.org/10.1109/BigData.2018.8621865  

   Huang, Rachel ; Pedoeem, Jonathan ; Chen, Cuixian ( December 2018 , 2018 IEEE International Conference on Big Data (Big Data), Seattle, WA, USA, 2018, pp. 2503-2510.)

   This paper focuses on YOLO-LITE, a real-time object detection model developed to run on portable devices such as a laptop or cellphone lacking a Graphics Processing Unit (GPU). The model was first trained on the PASCAL VOC dataset then on the COCO dataset, achieving a mAP of 33.81% and 12.26% respectively. YOLO-LITE runs at about 21 FPS on a non-GPU computer and 10 FPS after implemented onto a website with only 7 layers and 482 million FLOPS. This speed is 3.8 × faster than the fastest state of art model, SSD MobilenetvI. Based on the original object detection algorithm YOLOV2, YOLO-LITE was designed to create a smaller, faster, and more efficient model increasing the accessibility of real-time object detection to a variety of devices. 

   more » « less
4. Constraining Emissions of Volatile Organic Compounds Over the Indian Subcontinent Using Space‐Based Formaldehyde Measurements

   https://doi.org/10.1029/2019JD031262  

   Chaliyakunnel, Sreelekha ; Millet, Dylan B. ; Chen, Xin ( October 2019 , Journal of Geophysical Research: Atmospheres)

   India is an air pollution mortality hot spot, but regional emissions are poorly understood. We present a high‐resolution nested chemical transport model (GEOS‐Chem) simulation for the Indian subcontinent and use it to interpret formaldehyde (HCHO) observations from two satellite sensors (OMI and GOME‐2A) in terms of constraints on regional volatile organic compound (VOC) emissions. We find modeled biogenic VOC emissions to be overestimated by ~30–60% for most locations and seasons, and derive a best estimate biogenic flux of 16 Tg C/year subcontinent‐wide for year 2009. Terrestrial vegetation provides approximately half the total VOC flux in our base‐case inversions (full uncertainty range: 44–65%). This differs from prior understanding, in which biogenic emissions represent >70% of the total. Our derived anthropogenic VOC emissions increase slightly (13–16% in the base case, for a subcontinent total of 15 Tg C/year in 2009) over RETRO year 2000 values, with some larger regional discrepancies. The optimized anthropogenic emissions agree well with the more recent CEDS inventory, both subcontinent‐wide (within 2%) and regionally. An exception is the Indo‐Gangetic Plain, where we find an underestimate for both RETRO and CEDS. Anthropogenic emissions thus constitute 37–50% of the annual regional VOC source in our base‐case inversions and exceed biogenic emissions over the Indo‐Gangetic Plain, West India, and South India, and over the entire subcontinent during winter and post‐monsoon. Fires are a minor fraction (<7%) of the total regional VOC source in the prior and optimized model. However, evidence suggests that VOC emissions in the fire inventory used here (GFEDv4) are too low over the Indian subcontinent.

    

   more » « less
5. Algorithm-hardware Co-optimization for Energy-efficient Drone Detection on Resource-constrained FPGA

   https://doi.org/10.1145/3583074  

   Suh, Han-Sok ; Meng, Jian ; Nguyen, Ty ; Kumar, Vijay ; Cao, Yu ; Seo, Jae-Sun ( June 2023 , ACM Transactions on Reconfigurable Technology and Systems)

   Convolutional neural network (CNN)-based object detection has achieved very high accuracy; e.g., single-shot multi-box detectors (SSDs) can efficiently detect and localize various objects in an input image. However, they require a high amount of computation and memory storage, which makes it difficult to perform efficient inference on resource-constrained hardware devices such as drones or unmanned aerial vehicles (UAVs). Drone/UAV detection is an important task for applications including surveillance, defense, and multi-drone self-localization and formation control. In this article, we designed and co-optimized an algorithm and hardware for energy-efficient drone detection on resource-constrained FPGA devices. We trained an SSD object detection algorithm with a custom drone dataset. For inference, we employed low-precision quantization and adapted the width of the SSD CNN model. To improve throughput, we use dual-data rate operations for DSPs to effectively double the throughput with limited DSP counts. For different SSD algorithm models, we analyze accuracy or mean average precision (mAP) and evaluate the corresponding FPGA hardware utilization, DRAM communication, and throughput optimization. We evaluated the FPGA hardware for a custom drone dataset, Pascal VOC, and COCO2017. Our proposed design achieves a high mAP of 88.42% on the multi-drone dataset, with a high energy efficiency of 79 GOPS/W and throughput of 158 GOPS using the Xilinx Zynq ZU3EG FPGA device on the Open Vision Computer version 3 (OVC3) platform. Our design achieves 1.1 to 8.7× higher energy efficiency than prior works that used the same Pascal VOC dataset, using the same FPGA device, but at a low-power consumption of 2.54 W. For the COCO dataset, our MobileNet-V1 implementation achieved an mAP of 16.8, and 4.9 FPS/W for energy-efficiency, which is ∼ 1.9× higher than prior FPGA works or other commercial hardware platforms.

    

   more » « less