Getting Started with Object Detection (ML) in Block Coding | PictoBlox Machine Learning Environment

hello everyone welcome to the pictor blocks machine learning environment tutorial Series in this tutorial we're going to learn about object detection object detection is one of the machine learning model types which can be trained in pictor blocks object detection can be used to locate and recognize objects in a given image for example an object detection model trained to recognize alarm clocks will locate alarm clocks in an image using bounding box es in this example we're going to use object detection to detect bottles and ey classes in a given image however you can pick any two objects of your choice the two classes we'll be working with are bottle and glasses let's get started open pictor blocks and select the block coding environment go to the machine learning environment by selecting the open ml environment option under the files tab as we are training our models in Python it is important that we have the required dependencies in order to download these dependencies simply click on the gear icon on the top right of your screen and select the download dependencies option this will download and update the dependencies required to train the model click on the create new project option to initialize your project type an appropriate name of the project and select object detection as the model type click on the create project button and you'll see the object detection window you'll find yourself under the import tab with three options to import images by using the webcam by selecting the images from the computer's hard drive and by downloading the images from a repository for this example we will be using our webcam to capture images you need to capture more than 25 images for object detection to work you'll be able to see all your captured images here if you do not like any image hover with the image and click on the delete button to delete the image labeling the data is crucial in object detection go to the bbox tab and you'll see the create bounding box window click on the create box button and label the object as you label more objects class info will be updated accordingly to label an image belonging to a class already created click the edit button and simply click on the corresponding class clicking on the next button will automatic a Ally save your bonding box and label however make sure to press the save box button after you label the last [Music] image once the images are labeled go to the image tab to review the images once the images are reviewed go to the train tab this is where we will train the model click on the train new model button make sure all the classes are selected and click on the generate data set button once the data set is generated click on next in object detection we can work work with three hyper parameters bath size number of iterations and number of layers however over the hper parameters to view the description for this example we will be training the model for 5,000 iterations with a bath size of four we'll set the number of layers to four click on Create and then click on start draining object detection is a time-taking task so please be patient while training after the model is trained you can view the graphs by clicking on the corresponding buttons go to the evaluate tab to view a comprehensive overview of your model performance you can also view false positives and false negatives for your classes false positives are observations from a different class that the model detects to be in the current class false negatives are observations from the current class that the model detects to be in a different class now let's move to the test tab you can test your model here either by uploading the images or by using the webcam use the webcam and test your model you can alter the IOU and confidence threshold by adjusting the sliders on top to see the performance of the train model click on the pictor blocks option in the export model column and pictor blocks will load your model into the block coding environment observe how we have blocks for the model we just strained on the blocks pallet you can click on the open recognition window block and test the models working add the wind flag clicked block and the forever block into the scripting area and snap them together add a turn video on stage with transparency block above the forever block select on and zero as transparency this will make the camera feed show up on the stage add a set detection threshold to block and set the threshold to 0.5 follow it up with a bounding box block and set the drop down to show inside the for Block add a analyze image from block and set the drop down to camera now go to the variables pallet and create a variable named iteration add a set iteration to block and set the value to one we'll use the repeat block to recognize all the objects present in the frame drag a great number of objects block inside the repeat block satation space SP under the repeat block add a save block drag a class of object block and set it to the iteration variable follow a TP with a weight block of 1 second finally change the iteration variable by one run your script by clicking on the green flag to see the project in action there you have it an object detection project made in vctor blocks [Music] what

Channel: STEMpedia

Views: 7,388

Keywords:

Id: eqcPu8qs4EA

Length: 8min 40sec (520 seconds)

Published: Tue Nov 08 2022