An official website of the United States government
Efficient and adaptive computer vision systems have been proposed to make computer vision tasks, such as image classification and object detection, optimized for embedded or mobile devices. These solutions, quite recent in their origin, focus on optimizing the model (a deep neural network, DNN) or the system by designing an adaptive system with approximation knobs. Despite several recent efforts, we show that existing solutions suffer from two major drawbacks. First , while mobile devices or systems-on-chips (SOCs) usually come with limited resources including battery power, most systems do not consider the energy consumption of the models during inference. Second , they do not consider the interplay between the three metrics of interest in their configurations, namely, latency, accuracy, and energy. In this work, we propose an efficient and adaptive video object detection system — Virtuoso , which is jointly optimized for accuracy, energy efficiency, and latency. Underlying Virtuoso is a multi-branch execution kernel that is capable of running at different operating points in the accuracy-energy-latency axes, and a lightweight runtime scheduler to select the best fit execution branch to satisfy the user requirement. We position this work as a first step in understanding the suitability of various object detection kernels on embedded boards in the accuracy-latency-energy axes, opening the door for further development in solutions customized to embedded systems and for benchmarking such solutions. Virtuoso is able to achieve up to 286 FPS on the NVIDIA Jetson AGX Xavier board, which is up to 45 times faster than the baseline EfficientDet D3 and 15 times faster than the baseline EfficientDet D0. In addition, we also observe up to 97.2% energy reduction using Virtuoso compared to the baseline YOLO (v3) — a widely used object detector designed for mobiles. To fairly compare with Virtuoso , we benchmark 15 state-of-the-art or widely used protocols, including Faster R-CNN (FRCNN) [NeurIPS’15], YOLO v3 [CVPR’16], SSD [ECCV’16], EfficientDet [CVPR’20], SELSA [ICCV’19], MEGA [CVPR’20], REPP [IROS’20], FastAdapt [EMDL’21], and our in-house adaptive variants of FRCNN+, YOLO+, SSD+, and EfficientDet+ (our variants have enhanced efficiency for mobiles). With this comprehensive benchmark, Virtuoso has shown superiority to all the above protocols, leading the accuracy frontier at every efficiency level on NVIDIA Jetson mobile GPUs. Specifically, Virtuoso has achieved an accuracy of 63.9%, which is more than 10% higher than some of the popular object detection models, FRCNN at 51.1%, and YOLO at 49.5%.
more »
« less
YOLO-LITE: A Real-Time Object Detection Algorithm Optimized for Non-GPU Computers
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
- Award ID(s):
- 1659288
- PAR ID:
- 10132924
- Date Published:
- Journal Name:
- 2018 IEEE International Conference on Big Data (Big Data), Seattle, WA, USA, 2018, pp. 2503-2510.
- Page Range / eLocation ID:
- 2503 to 2510
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Deep learning models have significantly improved object detection, which is essential for visual sensing. However, their increasing complexity results in higher latency and resource consumption, making real-time object detection challenging. In order to address the challenge, we propose a new lightweight filtering method called L-filter to predict empty video frames that include no object of interest (e.g., vehicles) with high accuracy via hybrid time series analysis. L-filter drops those frames deemed empty and conducts object detection for nonempty frames only, significantly enhancing the frame processing rate and scalability of real-time object detection. Our evaluation demonstrates that L-filter improves the frame processing rate by 31–47% for a single traffic video stream compared to three standalone state-of-the-art object detection models without L-filter. Additionally, L-filter significantly enhances scalability; it can process up to six concurrent video streams in one commodity GPU, supporting over 57 fps per stream, by working alongside the fastest object detection model among the three models.more » « less
-
Recent advances in computer vision has led to a growth of interest in deploying visual analytics model on mobile devices. However, most mobile devices have limited computing power, which prohibits them from running large scale visual analytics neural networks. An emerging approach to solve this problem is to offload the computation of these neural networks to computing resources at an edge server. Efficient computation offloading requires optimizing the trade-off between multiple objectives including compressed data rate, analytics performance, and computation speed. In this work, we consider a “split computation” system to offload a part of the computation of the YOLO object detection model. We propose a learnable feature compression approach to compress the intermediate YOLO features with lightweight computation. We train the feature compression and decompression module together with the YOLO model to optimize the object detection accuracy under a rate constraint. Compared to baseline methods that apply either standard image compression or learned image compression at the mobile and perform image de-compression and YOLO at the edge, the proposed system achieves higher detection accuracy at the low to medium rate range. Furthermore, the proposed system requires substantially lower computation time on the mobile device with CPU only.more » « less
-
Artificial Intelligence (AI) developments in recent years have allowed several new types of applications to emerge. In particular, detecting people and objects from sequences of pictures or videos has been an exciting field of research. Even though there have been notable achievements with the emergence of sophisticated AI models, there needs to be a specialized research effort that helps people finding misplaced items from a set of video sequences. In this paper, we leverage voice recognition and Yolo (You Only Look Once) real-time object detection system to develop an AI-based solution that addresses this challenge. This solution assumes that previous recordings of the objects of interest and storing them in the dataset have already occurred. To find a misplaced object, the user delivers a voice command that is in turn fed into the Yolo model to detect where and when the searched object was seen last. The outcome of this process is a picture that is provided as evidence. We used Yolov7 for object detection thanks to its better accuracy and wider database while leveraging Google voice recognizer to translate the voice command into text. The initial results we obtained show a promising potential for the success of our approach. Our findings can be extended to be applied to various other scenarios ranging from detecting health risks for elderly people to assisting authorities in locating potential persons of interest.more » « less
-
Traffic intersections are prime locations for deployment of infrastructure sensors and edge computing nodes to realize the vision of a smart city. It is expected that the needs of a smart city, in regards to traffic and pedestrian traffic systems monitored by cameras/video, can be met by using stateof-the-art artificial-intelligence (AI) based object detectors and trackers. A critical component in designing an effective real-time object detection/tracking pipeline is the understanding of how object density, i.e., the number of objects in a scene, and imageresolution and frame rate influence the performance metrics. This study explores the accuracy and speed metrics with the goal of supporting pipelines that meet the precision and latency needs of a real-time environment. We examine the impact of varying image-resolution, frame rate and object-density on the object detection performance metrics. The experiments on the COSMOS testbed dataset show that varying the frame width from 416 pixels to 832 pixels, and cropping the images to a square resolution, result in the increase in average precision for all object classes. Decreasing the frame rate from 15 fps to 5 fps preserves more than 90% of the highest F1 score achieved for all object classes. The results inform the choice of video preprocessing stages, modifications to established AI-based object detection/tracking methods, and suggest optimal hyper-parameter values. Index Terms—Object Detection, Smart City, Video Resolution, Deep Learning Models.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1109/BigData.2018.8621865
