TY - GEN
T1 - Enabling incremental knowledge transfer for object detection at the edge
AU - Farhadi, Mohammad
AU - Ghasemi, Mehdi
AU - Vrudhula, Sarma
AU - Yang, Yezhou
N1 - Funding Information:
The National Science Foundation under the Robust Intelligence Program (#1750082) and I/UCRC Center for Embedded Systems (#1361926), the IoT Innovation (I-square) fund provided by ASU Fulton Schools of Engineering are gratefully acknowledged.
Funding Information:
In conclusion, we designed and implemented a frame-work for incremental knowledge transfer in edge computing environment. The parameters of a shallow model running on the user-end device are updated during inference at some key frames to achieve the close accuracy as using a deep model. We demonstrated the proposed approach in the real-world scenario. Our framework consisting of a shallow and a deep model resulted in 78% energy reduction when compared to running the deep model alone. The experiments also revealed that the latency of communication must be accounted for when deciding to do the model updates. Acknowledgment: The National Science Foundation under the Robust Intelligence Program (#1750082) and I/UCRC Center for Embedded Systems (#1361926), the IoT Innovation (I-square) fund provided by ASU Fulton Schools of Engineering are gratefully acknowledged.
Publisher Copyright:
© 2020 IEEE.
PY - 2020/6
Y1 - 2020/6
N2 - Object detection using deep neural networks (DNNs) involves a huge amount of computation which impedes its implementation on resource/energy-limited user-end devices. The reason for the success of DNNs is due to having knowledge over all different domains of observed environments. However, we need a limited knowledge of the observed environment at inference time which can be learned using a shallow neural network (SHNN). In this paper, a systemlevel design is proposed to improve the energy consumption of object detection on the user-end device. An SHNN is deployed on the user-end device to detect objects in the observing environment. Also, a knowledge transfer mechanism is implemented to update the SHNN model using the DNN knowledge when there is a change in the object domain. DNN knowledge can be obtained from a powerful edge device connected to the user-end device through LAN or Wi-Fi. Experiments demonstrate that the energy consumption of the user-end device and the inference time can be improved by 78% and 40% compared with running the deep model on the user-end device.
AB - Object detection using deep neural networks (DNNs) involves a huge amount of computation which impedes its implementation on resource/energy-limited user-end devices. The reason for the success of DNNs is due to having knowledge over all different domains of observed environments. However, we need a limited knowledge of the observed environment at inference time which can be learned using a shallow neural network (SHNN). In this paper, a systemlevel design is proposed to improve the energy consumption of object detection on the user-end device. An SHNN is deployed on the user-end device to detect objects in the observing environment. Also, a knowledge transfer mechanism is implemented to update the SHNN model using the DNN knowledge when there is a change in the object domain. DNN knowledge can be obtained from a powerful edge device connected to the user-end device through LAN or Wi-Fi. Experiments demonstrate that the energy consumption of the user-end device and the inference time can be improved by 78% and 40% compared with running the deep model on the user-end device.
UR - http://www.scopus.com/inward/record.url?scp=85090122373&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85090122373&partnerID=8YFLogxK
U2 - 10.1109/CVPRW50498.2020.00206
DO - 10.1109/CVPRW50498.2020.00206
M3 - Conference contribution
AN - SCOPUS:85090122373
T3 - IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops
SP - 1591
EP - 1599
BT - Proceedings - 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2020
PB - IEEE Computer Society
T2 - 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2020
Y2 - 14 June 2020 through 19 June 2020
ER -