search

weiliu89 / caffe

Caffe: a fast open framework for deep learning.
http://caffe.berkeleyvision.org/
Other
4.77k stars 1.68k forks source link

Frame skipping is not running in caffe SSD #827

Open NiranjanAntany opened 6 years ago

NiranjanAntany commented 6 years ago

HI Weiliu, Am very interested in deep learning frameworks, i tried SSD caffe in NVIDIA TX2, but i got only 8-9FPS , I need to speed the performance in NVIDIA TX2, I tried many methods like frame skipping in video_data_layer.cpp but it didn't worked, the data is not going to the layers after frame skipping, please tell me the method to speed up the process to achieve upto 20FPS with pretrained VOC model file only..

Am new to this please help me...

xhzzhang commented 6 years ago

Hello, i have trouble run it in TX2, got below message:

F0301 12:35:45.320708 14370 pooling_layer.cu:212] Check failed: error == cudaSuccess (8 vs. 0) invalid device function Check failure stack trace: @ 0x7f83914b94 (unknown) @ 0x7f83914aac (unknown) @ 0x7f83914500 (unknown) @ 0x7f8391717c (unknown) @ 0x7f83dea394 caffe::PoolingLayer<>::Forward_gpu() @ 0x7f83beab18 caffe::Net<>::ForwardFromTo() @ 0x7f83beae04 caffe::Net<>::Forward() @ 0x408dec test() @ 0x4074a0 main @ 0x7f8353de08 (unknown) Aborted

Did you have got same error? how do you fix it?