Now it's time to create a video classification model using Auto ML. So again remember exist under Vertex AI. And first of all we need to create a managed data set. So under Vertex AI under datasets, we want to create a new data set for this problem. In this case I will call video conchita dataset, which kind of data set we want to create? In this case we want to create a model that will be a video classification model. So this is why here I will select video classification and I will create this data set in the region, us-central1, I click create and now we need to upload the videos. All the videos contains images of different sports and the model will classify which kind of sport is present in the video. This video that we will use to train our model are all ready uploaded into a google cloud storage bucket. So this is why in the import method we need to select import files from Google Cloud storage, and we need to pass here the path to the bucket in which we have all the videos uploaded. So I specify here the path and I click in the continue button. So now again as always we need to wait few minutes while the important job is working. So now the system is uploading all these different videos in this managed data set to train a video classification model. Good, now our data set is all ready created. So you can see that we have 500 videos of different sports, we have 500 videos that belongs to five different classes of a sport. Time now to train our model. So we click under train a new model and again in this case we will use Auto ML model, Google will make all the hard work for us and click under Continue. We specify a name for this model. And I call this model video_conchita and I click the button, Start Training. And again, we need to wait in this case it's long time, we need to wait around three hours until the model job is finished to try to test and deploy our model to see how it works or send a batch prediction job to our model. Great, our model finish or Auto ML finished the training job to create our model. So here we can see the video_conchita under the model staff inside of Vertex AI. If I click over this video_conchita, we can see the performance metrics. We can evaluate also the confusion metrics to see if the performance metrics that we are obtaining are good enough for our business needs. If we are happy with the results, in this case, instead of deploying the model to an end point and test the model using online prediction job, we will run a batch prediction job. That is, we will pass a multiple amount of videos to the model and the model will need to make multiple predictions and give to us the results. So to do this job, I want to restore the results in a Google Cloud storage bucket that will exist under my project. So for that, let's first go and create a Google Cloud storage bucket under our project. So from here you can see that I am under my project and remember always the project ID that we have is unique. So this is why because I want to create a Cloud Storage bucket, remember that the name of the Google Cloud storage bucket needs to be unique worldwide. So this is why I will assign the same name as my project ID. So I am going to Google Cloud storage bucket, okay, and I will create bucket, I will assign the name for my bucket. And in this case I want to create a regional bucket and I will have the regional bucket created in the same region that I am training and creating our model. So let's click Create, and you can see here that our bucket is all ready created and is completely empty. When we send the batch job to the model, we want to store here the results for all the prediction jobs. So let's create this batch prediction job. So from here, from the model, I click under Batch Prediction. Okay, so I need to create a batch prediction but job I assign a name. So I want to create a bad job over this model, and I need to specify what is the source in which we have all the different videos that we want to send to the model to do their job. And you can see here that the format that the model expect is Jason L. So for that, we all ready have uploaded few videos and then a Google Cloud storage bucket in the format that the model expect. So this is why here, I will specify this source path in this bucket, I have all ready a lot multiple videos to send this batch prediction job. And here for the destination path, I will put [INAUDIBLE] a bucket that we just create a few seconds ago, so I will copy the name of the bucket, and I will create a folder that will be called predict results to store there our results. So, now I click Create button, and you can see that the match prediction job starts. And the status you spending. Now we are and the system is going to the bucket in which we have all the source videos and are sending these videos to the model to make the job to predict which kind of sport is present in each of these videos. The Auto ML model, the video recognition classification model is doing their job and we'll store the results in the storage bucket that we just create. So, let's wait a few minutes until we have this job done, and let's see the results, our batch job all ready finished. If we click here in the batch job, we can see that the model in this batch job, predict the results for five different inputs. And, here we have the export location in which the system automatically store the results. That will be the google cloud storage bucket that we create before. And because we send five different videos, we need to have now under this bucket five different results, five different predictions. Perfect, let's see if it's working good or not. Let's view the results. If I click here under View Results. Here we have a set of different videos, the five videos that we pass to the system and let's see for example, if the model is working fine, that is if it's giving to us good results. If I click under this video, the model predicts that the action that is happening here is a cartwheel. If I click for example here, I love soccer. So let's see if the model is working fine. Look at the video and yes, the model is predicting a kick ball action. So our model is working perfectly. And here we can see the five different jobs that has been done. The five different predictions that the model have done in our batch prediction job. So good work. You create an image, not me, image, sorry, you create a video recognition classification model using Auto ML. See you in the next lab.