Session Track
November 03 (Friday)
IT융합 전문가리뷰(I)/인공지능
Oral,
제11발표장(304호),
13:00~14:00
- 이덕진(군산대)
Fr11D-2
13:00~14:00
Embedded Computer Vision with Nvidia TX2
This paper deploys the deep learning for the live image recognition and obstacle detection. It tries to solve the computer
vision problems of recognition, detection and segmentation using the present days deep learning methods (CNN).
It uses the models from GoogleNet and AlexaNet for live image recognition. The deployment is into the NVidia Jetson
TX1 embedded GPU, which uses its default camera for the recognition. The work uses an Ubuntu-16.04 based PC with
GTX-1080 GPU as a server for the custom model training. Cuda-8.0, Cudnn-5.1 and NVidia DIGITS library; are used
in the training process. The model is deployed into the embedded Nvidia Jetson TX1 and operated in the real-time image
recognition.
vision problems of recognition, detection and segmentation using the present days deep learning methods (CNN).
It uses the models from GoogleNet and AlexaNet for live image recognition. The deployment is into the NVidia Jetson
TX1 embedded GPU, which uses its default camera for the recognition. The work uses an Ubuntu-16.04 based PC with
GTX-1080 GPU as a server for the custom model training. Cuda-8.0, Cudnn-5.1 and NVidia DIGITS library; are used
in the training process. The model is deployed into the embedded Nvidia Jetson TX1 and operated in the real-time image
recognition.
Keywords : Deep Learning, GoogleNet, AlexaNet, Nvidia-Jetson, CNN
Paper : Fr11D-2.pdf
(사)대한기계학회, 서울시 강남구 테헤란로 7길 22 한국과학기술회관 신관 702호, Tel: (02)501-3646~3648, E-mail: ksme@ksme.or.kr