Revision history [back]

Trying to do object detection by Halide(tensorflow ssd_mobilenet_v1_coco ), before I set Halide as backend, fps is around 22~23, cpu rate is around 70~80%, memory usage of gpu ram is around 431MiB(monitor by nvidia-smi)

After I turned on Halide,

net.setPreferableBackend(cv::dnn::DNN_BACKEND_HALIDE);

fps drop down to 6~10fps, usage of cpu and gpu ram still similar to the case without Halide.

I intent to leverage gpu power to speed things up and lower the burden of cpu, but looks like Halide cannot speed things up nor lower the cpu usage.Any suggestion to speed things up by Halide? Thanks

Trying to do object detection by Halide(tensorflow ssd_mobilenet_v1_coco ), before I set Halide as backend, fps is around 22~23, cpu rate is around 70~80%, memory usage of gpu ram is around 431MiB(monitor by nvidia-smi)

After I turned on Halide,

net.setPreferableBackend(cv::dnn::DNN_BACKEND_HALIDE);

fps drop down to 6~10fps, usage of cpu and gpu ram still similar to the case without Halide.

I intent to leverage gpu power to speed things up and lower the burden of cpu, but looks like Halide cannot speed things up nor lower the cpu usage.Any suggestion to speed things up by Halide? usage.Am I doing anything wrong? Thanks

Trying to do object detection by Halide(tensorflow ssd_mobilenet_v1_coco ), before I set Halide as backend, fps is around 22~23, cpu rate is around 70~80%, memory usage of gpu ram is around 431MiB(monitor by nvidia-smi)

After I turned on Halide,

net.setPreferableBackend(cv::dnn::DNN_BACKEND_HALIDE);

fps drop down to 6~10fps, usage of cpu and gpu ram still similar to the case without Halide.

I intent to leverage gpu power to speed things up and lower the burden of cpu, but looks like Halide cannot speed things up nor lower the cpu usage.Am I doing anything wrong? Thanks

Trying to do object detection by Halide(tensorflow ssd_mobilenet_v1_coco ), Halide, before I set Halide as backend, fps is around 22~23, cpu rate is around 70~80%, memory usage of gpu ram is around 431MiB(monitor by nvidia-smi)

After I turned on Halide,

net.setPreferableBackend(cv::dnn::DNN_BACKEND_HALIDE);

fps drop down to 6~10fps, usage of cpu and gpu ram still similar to the case without Halide.

I intent to leverage gpu power to speed things up and lower the burden of cpu, but looks like Halide cannot speed things up nor lower the cpu usage.Am I doing anything wrong? Thanks

Trying to do object detection by Halide, before I set Halide as backend, fps is around 22~23, cpu rate is around 70~80%, memory usage of gpu ram is around 431MiB(monitor by nvidia-smi)

After I turned on Halide, Halide(every codes remain the same),

net.setPreferableBackend(cv::dnn::DNN_BACKEND_HALIDE);

fps drop down to 6~10fps, usage of cpu and gpu ram still similar to the case without Halide.

I intent to leverage gpu power to speed things up and lower the burden of cpu, but looks like Halide cannot speed things up nor lower the cpu usage.Am I doing anything wrong? Thanks

Trying to do object detection by Halide, before I set Halide as backend, fps is around 22~23, cpu rate is around 70~80%, memory usage of gpu ram is around 431MiB(monitor by nvidia-smi)

After I turned on Halide(every codes remain the same),

net.setPreferableBackend(cv::dnn::DNN_BACKEND_HALIDE);

fps drop down to 6~10fps, usage of cpu and gpu ram still similar to the case without Halide.

I intent to leverage gpu power to speed things up and lower the burden of cpu, but looks like Halide cannot speed things up nor lower the cpu usage.Am I doing anything wrong? wrong(should I process multiple batch at the same time)? Thanks

Trying to do object detection by Halide, before I set Halide as backend, fps is around 22~23, cpu rate is around 70~80%, memory usage of gpu ram is around 431MiB(monitor by nvidia-smi)

After I turned on Halide(every codes remain the same),

net.setPreferableBackend(cv::dnn::DNN_BACKEND_HALIDE);

fps drop down to 6~10fps, usage of cpu and gpu ram still similar to the case without Halide.

I intent to leverage gpu power to speed things up and lower the burden of cpu, but looks like Halide cannot speed things up nor lower the cpu usage.Am I doing anything wrong(should wrong? should I process multiple batch at the same time)? Thankstime? No matter what, it is weird that Halide occupy so many cpu resource, I though most of the heavy tasks should be handle by gpu.

Trying to do object detection by Halide, before I set Halide as backend, fps is around 22~23, cpu rate is around 70~80%, memory usage of gpu ram is around 431MiB(monitor by nvidia-smi)

After I turned on Halide(every codes remain the same),

net.setPreferableBackend(cv::dnn::DNN_BACKEND_HALIDE);

fps drop down to 6~10fps, usage of cpu and gpu ram still similar to the case without Halide.

I intent to leverage gpu power to speed things up and lower the burden of cpu, but looks like Halide cannot speed things up nor lower the cpu usage.Am I doing anything wrong? should I process multiple batch at the same time? No matter what, it is weird that Halide occupy so many cpu resource, I though most of the heavy tasks should be handle by gpu.

Trying to do object detection by Halide, before I set Halide as backend, fps is around 22~23, cpu rate is around 70~80%, memory usage of gpu ram is around 431MiB(monitor by nvidia-smi)

After I turned on Halide(every codes remain the same), same)

net.setPreferableBackend(cv::dnn::DNN_BACKEND_HALIDE);

fps drop down to 6~10fps, usage of cpu and gpu ram still similar to the case without Halide.

At first I intent to leverage gpu power to speed things up and lower the burden of cpu, but looks like Halide cannot speed things up nor lower the cpu usage.Am I doing anything wrong? should I process multiple batch at the same time? No matter what, think it is weird that Halide occupy because the network and batch size are too small, so many cpu resource, I though most of increase the heavy tasks should be handle by gpu.

Edit : I edit the codes batch size from 1 to process 32 batch at once, 32. However, the usage of gpu ram do not increase at all, all even I increase the batch size to 32, instead cpu usage raise over 90%, looks like Halide backend never use gpu at allall.

Trying to do object detection by Halide, before I set Halide as backend, fps is around 22~23, cpu rate is around 70~80%, memory usage of gpu ram is around 431MiB(monitor by nvidia-smi)

After I turned on Halide(every codes remain the same)

net.setPreferableBackend(cv::dnn::DNN_BACKEND_HALIDE);

fps drop down to 6~10fps, usage of cpu and gpu ram still similar to the case without Halide.

At first I think it is because the network and batch size are too small, so I increase the batch size from 1 to 32. However, the usage of gpu ram do not increase at all even I increase the batch size to 32, instead cpu usage raise over 90%, looks like Halide backend never use gpu at all.

Trying to do object detection by Halide, before I set Halide as backend, fps is around 22~23, cpu rate is around 70~80%, memory usage of gpu ram is around 431MiB(monitor by nvidia-smi)

After I turned on Halide(every codes remain the same)

net.setPreferableBackend(cv::dnn::DNN_BACKEND_HALIDE);

fps drop down to 6~10fps, usage of cpu and gpu ram still similar to the case without Halide.

At first I think it is because the network and batch size are too small, so I increase the batch size from 1 to 32. However, the usage of gpu ram do not increase at all even I increase the batch size to 32, instead cpu usage raise over 90%, looks like Halide backend never use gpu at all.

Trying to do object detection by Halide, before I set Halide as backend, fps is around 22~23, cpu rate is around 70~80%, memory usage of gpu ram is around 431MiB(monitor by nvidia-smi)

After I turned on Halide(every codes remain the same)

net.setPreferableBackend(cv::dnn::DNN_BACKEND_HALIDE);

fps drop down to 6~10fps, usage of cpu and gpu ram still similar to the case without Halide.

At first I think it is because the network and batch size are too small, so I increase the batch size from 1 to 32. However, the usage of gpu ram do not increase at all even I increase the batch size to 32, instead cpu usage raise over 90%, looks like Halide backend never use gpu at all.