Subtitles section Play video
-
Welcome back
-
I'd like to show you the full flow of inferencing with OpenVINO
-
Thru the flow you'll be able to see the various utilities that are included in the OpenVINO
-
package. Let's begin
-
In a very high level
-
We want to start with a DL based model that is already trained.
-
We want to prepare it for inference
-
We want to implement the inference in our application
-
We want the inference to run on multiple devices CPU, GPU, VPU or FPGA
-
And we might even want to implement custom layers per each of these devices..
-
And now in a little more details.
-
the assumption is that you already have a trained model.
-
The model is usually described by few files, .pb file for TensorFlow
-
.params for Mxnet and so on. Here I'm showing a Caffe model file,
-
Connectivity is described by the prototex.. weights and biases are in the .caffemodel
-
file In Video #6, we'll show you how to easily
-
use our model-downloader to get a public model.
-
Now we have to prepare this model for inference, we need to optimize it, change the data format
-
if needed etc. All of these tasks are performed by the Model-Optimizer
-
Check out Video 8 for an explanation on the Model Optimizer concept
-
In Video 9, we'll demonstrate the basic functions, video 10 more advance ones and so on
-
Now that we have the model ready for inference, we are using the "Inference Engine" API
-
in our application to actually do the inference Check out video 15 for a short explanation
-
of the concept, In Video 16 we'll show the API using a sample
-
code and so on..
-
On Video 19 we'll show you how easy it is to run inference on difference devices with
-
the Inference engine, You only have to code once, and you can run
-
the same model on a CPU or a GPU or many other devices which is really cool
-
And if you want to code your custom layers for CPU or GPU,
-
We'll show you how to do it on Videos 25 and 26..
-
So this is the full flow, it is simple and we will go over each stage in details.
-
In coming videos..
-
in the next video let's just try to run a demo that demonstrate the full flow..
-
Subscribe to out channel and we'll send you a new video each time we have a new feature
-
of OpenVINO..
-
Thank you.