On this video, I discuss the start of planning a workflow of how I think the models will collect data, train, and test models.
Welcome to my machine learning blog on the next series of videos I will be discussing how I plan to use machine learning framework with other technology frameworks to create applications please join me by subscribing to my blog and ask questions or give answers to the problems I encounter on the concepts I am building my guys since it started in on so I been looking at soon as you year when do the bills a flowchart and really while he flowchart is to make sure keeps is in line with his winnings so well I do first is review eggs workflow here so steps one to bring must describe the installation of long been in a Windows environment we went to the floor and one of the experiment sister says I’m not focused on a solo the next cashier that he has is gathering in labeling pictures and then That is actually generating training data and then creating a label map the configuring training: the extent of that is to actually execute training in intensive long ones that training days it once a training is completed we then exported into a frozen in France and friends grass and then we do a testing of that data and knowledge to take so steps three through eight is without focus on this particular projects so our first step year is to identify was behind what so my case here I guess we can identify the user/data source right came and then mentality video and no training McAuliffe workflow suddenly other function here is it’s attentively so it when a representative local becomes a flow the application the core application itself and including the model the training model that will be using following to this one so well into his party and swimming) but for now so to start this on we have to identify all of all our data source began and then from the here and Elizabeth Elizabeth Seabiscuit. It is on the extent after that is we say here gathering the data and then you start labeling pictures so the next step here will be labeling an open here classified classification so let labeling is is a process that would you do to create bonding boxes but we also have to identifying our classifications are classes so the classifications of browser was identify where classes are being sold in the argument that for all to simplify this this is one of the same song into the flow we use classes outside to the flow: labeling settlement but this both here in response to the next step every have he says here is generating training data to the generating training data is afterward don’t I labeling with you to create some kind of output that data that we created as part of the labeling process that the data up with his baby massage in a format that is a training data the to flow needs its 100 and create these two because again this training needed that training data and label map cremations to me his old ones that England during the instead of separately for his tutorial but biggest because it was down to configuring training and again that’s creating a little math and training so at however only the hearing is be putting this in this swimmingly because this can happen with intensive configure training is going to happen with intensive full of fish and so an extent he has here is the actual training is executed and exporting the reference still those tools are individual steps to happened separately and then it’s a sheer testing fight song when a step you here is execute training is will happen with intensive flow and then the next step year is to export training data set in the beginning of this one because there’s there’s two formats right on his reference for his particular tutorial and review the same thing right test this exporting and in France graph so there will be an additional step in with my adding here because depending on on an API that we use in our case the client kind site use so someone can you guys are as a sorry so this exporting references I’m not all adding nearby to put export training the so when when the training gets executed to explore that all of the files is created not been exported gardening is a frozen inference and then after that overdue is that passes to the user which of be the kind site and then here we to do test project And then once this is done a data set is done boldly doing is on within the transition into another phrase so that would certainly that will be testing our object detection models and to were satisfied that we have high confidence level is detecting the objects and wanted to detect and then once that happens than will transition to a point available for consumption or in production environment Nikes… Back to it’s a full serving maybe that’s a full serving for for sure tear gas or Juliet different different APIs and then on at the end after that them go back to the client site we build APIs will have a UI for interfacing and that pretty much will end on workflow here so you see here this pretty much been have as far as the machine learning portion is identify our data sources labeling classifications configure training execute training export training data sets test our object detection models and then Ted then be reused it for flaws serving may be used year it would exported to tear the JS which is Java scripting for the could be using a browser or TF mobile so they can use it into a smart phone mobile application android etc. etc. and of course the UI will be a user interface where inorganic available application web application a standalone application like of the robot or something like that in that will be the end of our data when machine learning from start to finish so it’s your project go to flow this way the biggest thing that I want to added to this flow is when we get into specifics on how to get a data source so one of the things see tonight insert the to the city of history of projects right for your first patriots at a house a high-level overview is snuggle with wanted to do that in anyways is features with features on half general to specific zone patriots well wanted to this build a web interface so what data sources that have is I personally admonish on the YouTube video where I have a dashboard camera on a car and I am driving around capturing the so that it one of my one of my use of sources is video right so what I want to have I want to upload and video somewhere forest the video and said they frames where the American flag see so so the reason I’m doing only a – a desk Is the desk Nine that are inches) is kind that’s To have this family gets mounted on my windshield a return on the night and I drive to work to the office is am driving through the office England to the road and stuff a Sure maybe some houses that have flagpoles and some buildings some commercial buildings that have the American flag industry and Mary like maybe like apartment complexes car sales lots so what I’m doing this as I’m driving this this – can is recording this video and that is creating one on my data sets for this video so I want to be able to overlook this video somewhere and then once I overlook this video I can parse the video and then extract the frames where I could see the American flag of this frames will eventually I’ll export amount and then I’ll start doing the labeling process so the next source that I have is used YouTube videos there is a robust so there’s there’s YouTube video channels where people you actually have like hours to hours of the American flag flapping around and that’s that that is another potential data source for this salt that’s where I’ll be getting the data set as well now there’s there’s this just my thoughts here Zafar the American flag is to capture the American flag flying during seasons. With spring summer fall and winter now I’m here in the South thing is that it doesn’t snowmachine some hoping that I’ll find somebody that has – cam up as a northeast when it’s snowing and somebody still left the American flag out there and they capture than the submitted so what is features that I want to have again is for and user to be able to upload a video for submission to support this project so again on behalf of end-users this might be one subscribers and the lives of their that wants to contribute to the to the data inputs in their upload a video only parses year and New England execute to training so what are the other things I want to have release for the patriots is in of the to different flags undeniably have 50 states there are no yet invalid the states now but to also be able to detect place in general right now mean there’s there’s 50 states each states have fits has its own flag but the ability to training those as well but in general maybe just able to detect in a flag of be interesting to some desk can YouTube videos living a user has or less than so teleported get web interface and in the to detect additional now the next feature that want to have is once once they know this is the have on your phrase to be able to take a picture with lives long spell be the latitude and longitude in the kitchen with a picture was thinking and a direction so give me knotting the picture this picture the submission will only use data set to validate is the flag exists on the picture so this will prevent sailors of the flag is not there is not detected that picture most likely was good be discarded also is automating validation of that flag exists in the exists then will be use the data set the addition they said. Of the Lancelot in the direction this pinpoints the location on a map but when the things that I want to do is there’s the this Google Chrome experiment where has a goal-will it is good to be around to try to download the data set I like to build this feature into this but repetition so people can see where we are you selling tickets he shows with patriots are so let’s say here is that than this one and it’s going to open the world map here and it shows on the pixels he will television industry is a map world innovation right so what I would like to show lists is this United States and shows where all my submissions I just to do is just the facts it to say project I like to do is a personal project so were patriots are alpha little at all the Medicare So Interdependence Patriots He> so Once Said Was That the Nine. Data to Create the World Patriots so That the Some of the Things That Don’t Have Patriots so the Next Project Is My Operations Investigations so Pedantic Posse and It Is Significantly More Details and Then Will Continue a Bit of Disclaimer Here I’ll Post of His Video Blog Are My Personal Opinion and They Don’t in Any Way Reflect Opinions of My Order All Materials Posts and Advice from This Site Are Informational Researching and for Testing Purposes of You Can Use Them at Your Own Responsibilities I Not in Any Way Responsible for Any Damage Done by Following Posts Advice Tutorials and Articles from This Blog My Machine Learning Blog May Not Own Some of the Continent Percent