Low power real-time object detection is an interesting application in deep learning with applications in smart wearables, Advanced Driver Assistance Systems (ADAS), drone surveillance systems, etc. In this paper, we discuss the limitations with existing networks and enumerate the various factors to keep in mind while designing neural networks for a target hardware. Based on our experience of working with TI embedded platform, we provide a systematic approach for designing real time object detection networks on low power embedded platforms. First stage involves identifying the optimal layers for the hardware, by understanding it's computational and memory limitations. The next step is to use these layers to come up with a basic building block that has low computational complexity. The final stage involves using model compression techniques like sparsification/quantization to accelerate the inference process. Based on this design approach, we were able to come up with a low latency object detection model HX-LPNet that operates at 22 FPS on low power TDA2PX System on Chip(SoC) provided by Texas Instruments (TI).
Description
Real-Time Object Detection On Low Power Embedded Platforms - IEEE Conference Publication
%0 Conference Paper
%1 9021990
%A Jose, G.
%A Kumar, A.
%A Kruthiventi S S, S.
%A Saha, S.
%A Muralidhara, H.
%B 2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW)
%D 2019
%K order1 real-time
%P 2485-2492
%R 10.1109/ICCVW.2019.00304
%T Real-Time Object Detection On Low Power Embedded Platforms
%U https://ieeexplore.ieee.org/document/9021990
%X Low power real-time object detection is an interesting application in deep learning with applications in smart wearables, Advanced Driver Assistance Systems (ADAS), drone surveillance systems, etc. In this paper, we discuss the limitations with existing networks and enumerate the various factors to keep in mind while designing neural networks for a target hardware. Based on our experience of working with TI embedded platform, we provide a systematic approach for designing real time object detection networks on low power embedded platforms. First stage involves identifying the optimal layers for the hardware, by understanding it's computational and memory limitations. The next step is to use these layers to come up with a basic building block that has low computational complexity. The final stage involves using model compression techniques like sparsification/quantization to accelerate the inference process. Based on this design approach, we were able to come up with a low latency object detection model HX-LPNet that operates at 22 FPS on low power TDA2PX System on Chip(SoC) provided by Texas Instruments (TI).
@inproceedings{9021990,
abstract = {Low power real-time object detection is an interesting application in deep learning with applications in smart wearables, Advanced Driver Assistance Systems (ADAS), drone surveillance systems, etc. In this paper, we discuss the limitations with existing networks and enumerate the various factors to keep in mind while designing neural networks for a target hardware. Based on our experience of working with TI embedded platform, we provide a systematic approach for designing real time object detection networks on low power embedded platforms. First stage involves identifying the optimal layers for the hardware, by understanding it's computational and memory limitations. The next step is to use these layers to come up with a basic building block that has low computational complexity. The final stage involves using model compression techniques like sparsification/quantization to accelerate the inference process. Based on this design approach, we were able to come up with a low latency object detection model HX-LPNet that operates at 22 FPS on low power TDA2PX System on Chip(SoC) provided by Texas Instruments (TI).},
added-at = {2020-04-21T20:33:09.000+0200},
author = {{Jose}, G. and {Kumar}, A. and {Kruthiventi S S}, S. and {Saha}, S. and {Muralidhara}, H.},
biburl = {https://www.bibsonomy.org/bibtex/2876ae0698777b3f6b90105607ea57219/sohnki},
booktitle = {2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW)},
description = {Real-Time Object Detection On Low Power Embedded Platforms - IEEE Conference Publication},
doi = {10.1109/ICCVW.2019.00304},
interhash = {6773872edd5b87a078627051edc83d3f},
intrahash = {876ae0698777b3f6b90105607ea57219},
issn = {2473-9944},
keywords = {order1 real-time},
month = oct,
pages = {2485-2492},
timestamp = {2020-06-02T20:00:07.000+0200},
title = {Real-Time Object Detection On Low Power Embedded Platforms},
url = {https://ieeexplore.ieee.org/document/9021990},
year = 2019
}