eduzhai > Applied Sciences > Engineering >

RoadNet-RT High Throughput CNN Architecture and SoC Design for Real-Time Road Segmentation

  • Save

... pages left unread,continue reading

Document pages: 10 pages

Abstract: In recent years, convolutional neural network has gained popularity in manyengineering applications especially for computer vision. In order to achievebetter performance, often more complex structures and advanced operations areincorporated into the neural networks, which results very long inference time.For time-critical tasks such as autonomous driving and virtual reality,real-time processing is fundamental. In order to reach real-time process speed,a light-weight, high-throughput CNN architecture namely RoadNet-RT is proposedfor road segmentation in this paper. It achieves 90.33 MaxF score on test setof KITTI road segmentation task and 8 ms per frame when running on GTX 1080GPU. Comparing to the state-of-the-art network, RoadNet-RT speeds up theinference time by a factor of 20 at the cost of only 6.2 accuracy loss. Forhardware design optimization, several techniques such as depthwise separableconvolution and non-uniformed kernel size convolution are customized designedto further reduce the processing time. The proposed CNN architecture has beensuccessfully implemented on an FPGA ZCU102 MPSoC platform that achieves thecomputation capability of 83.05 GOPS. The system throughput reaches 327.9frames per second with image size 1216x176.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×