This sample demonstrates how to build and execute inference in pipelined mode on example of classifications networks.
The pipelined mode might increase the throughput of the pictures. The latency of one inference will be the same as for synchronous execution.
The throughput increases due to follow reasons:
When two or more devices process one image, creating several infer requests and starting asynchronous inference allow for using devices in the most efficient way. If two devices are involved in execution, the most optimal value for -nireq
option is 2. To process infer requests more efficiently, Classification Sample Async uses round-robin algorithm. It starts execution of the current infer request and switches to waiting for results of the previous one. After finishing of waiting, it switches infer requests and repeat the procedure.
Another required aspect of good throughput is a number of iterations. Only with big number of iterations you can emulate the real application work and get good performance.
The batch mode is an independent attribute on the pipelined mode. Pipelined mode works efficiently with any batch size.
Upon the start-up, the sample application reads command line parameters and loads a network and an image to the Inference Engine plugin. Then application creates several infer requests pointed in -nireq
parameter and loads images for inference.
Then in a loop it starts inference for the current infer request and switches to waiting for the previous one. When results are ready, it swaps infer requests.
When inference is done, the application outputs data to the standard output stream.
NOTE: By default, Inference Engine samples and demos expect input with BGR channels order. If you trained your model to work with RGB order, you need to manually rearrange the default channels order in the sample or demo application or reconvert your model using the Model Optimizer tool with
--reverse_input_channels
argument specified. For more information about the argument, refer to When to Reverse Input Channels section of Converting a Model Using General Conversion Parameters.
Running the application with the -h
option yields the following usage message:
The command yields the following usage message:
Running the application with the empty list of options yields the usage message given above and an error message.
To run the sample, you can use AlexNet and GoogLeNet or other image classification models. You can download the pre-trained models with the OpenVINO Model Downloader or from https://download.01.org/opencv/.
NOTE: Before running the sample with a trained model, make sure the model is converted to the Inference Engine format (*.xml + *.bin) using the Model Optimizer tool.
You can do inference on an image using a trained AlexNet network on FPGA with fallback to CPU using the following command:
By default, the application outputs top-10 inference results for each infer request. It also provides throughput value measured in frames per seconds.