This demo showcases Object Detection with SSD and new Async API. Async API usage can improve overall frame-rate of the application, because rather than wait for inference to complete, the app can continue doing things on the host, while accelerator is busy. Specifically, this demo keeps two parallel infer requests and while the current is processed, the input frame for the next is being captured. This essentially hides the latency of capturing, so that the overall framerate is rather determined by the MAXIMUM(detection time, input capturing time) and not the SUM(detection time, input capturing time).
The technique can be generalized to any available parallel slack, e.g. doing inference and simultaneously encoding the resulting (previous) frames or running further inference, like some emotion detection on top of the face detection results, etc. There are important performance caveats though, for example the tasks that run in parallel should try to avoid oversubscribing the shared compute resources. E.g. if the inference is performed on the FPGA, and the CPU is essentially idle, than it makes sense to do things on the CPU in parallel. But if the inference is performed say on the GPU, than it can take little gain to do the (resulting video) encoding on the same GPU in parallel, because the device is already busy.
This and other performance implications and tips for the Async API are covered in the Optimization Guide
Other demo objectives are:
On the start-up the application reads command line parameters and loads a network to the Inference Engine. Upon getting a frame from the OpenCV's VideoCapture it performs inference and displays the results.
New "Async API" operates with new notion of the "Infer Request" that encapsulates the inputs/outputs and separates scheduling and waiting for result, next section. And here what makes the performance look different:
In this release, the Inference Engine also offers new API based on the notion of Infer Requests. One specific usability upside is that the requests encapsulate the inputs and outputs allocation, so you just need to access the blob with GetBlob method.
More importantly, you can execute a request asynchronously (in the background) and wait until ready, when the result is actually needed. In a mean time your app can continue :
Notice that there is no direct way to measure execution time of the infer request that is running asynchronously, unless you measure the Wait executed immediately after the StartAsync. But this essentially would mean the serialization and synchronous execution. This is what demo does for the default "SYNC" mode and reports as the "Detection time/fps" message on the screen. In the truly asynchronous ("ASYNC") mode the host continues execution in the master thread, in parallel to the infer request. And if the request is completed earlier than the Wait is called in the main thread (i.e. earlier than OpenCV decoded a new frame), that reporting the time between StartAsync and Wait would obviously incorrect. That is why in the "ASYNC" mode the inference speed is not reported.
Running the application with the
-h option yields the following usage message:
Running the application with the empty list of options yields the usage message given above and an error message. You can use the following command to do inference on GPU with a pre-trained object detection model:
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.
The only GUI knob is using 'Tab' to switch between the synchronized execution and the true Async mode.
The demo uses OpenCV to display the resulting frame with detections (rendered as bounding boxes and labels, if provided). In the default mode the demo reports