We all know how drones can be one of the most helpful objects in many things. I mean they can do a lot of things and make jobs easier, all thanks to the latest technology which keeps getting better day by day and is giving us major goals. Drones too are its creations, and when it comes to that, we expect nothing less.
Even though drones can be used for so many things which makes life easier, did you know that drones can also be used to recognize an apple? Yes, it is an apple we’re talking about.
Now it may sound really easy when we talk about it; the reality, however, is quite different; it is not so easy and simple at all. At least not for a drone which needs to constantly train its AI models into recognizing many pictures of apples. It could be an apple in different scenarios; it can be an apple in the shade, an apple in the bright sun, an apple from the front, or how an apple looks like from the back. How an apple looks like from a baby’s perspective or how an apple looks like, well from a frog’s perspective. This may sound silly to many of you right here, but did you know that machine learning also works in the very same way as we humans do, in other words, the machine also learns things the way we do with the same or perhaps more level of difficulty.
Let’s take an example of a baby; a baby tends to learn in many ways what an apple looks like. For a baby to completely realize what an apple looks like, it has to look at it for quite some time; it also has to go through a few pictures before it finally realizes that this is an apple and gets used to it. This theory was described by Andy Kong, who is a research assistant at the University of Hong Kong.
Now we know what might be going on in your minds at this very moment. It will be something like, oh wait, why are we talking about apples? Seriously apples?
Why are we even teaching drones to recognize apples? Well, the answer is simple; it is not just about apples, the whole idea is based on building up models for visual recognition which can be useful and can be effectively applied in many other ways.
Both Microsoft, as well as DJI, have a similar level of understanding that those drones that are powered with this type of technology can prove to be very effective in a certain type of scenarios. This is why Microsoft and DJI have teamed up to work on AI Drones that can recognize objects, Let’s say if you own a huge warehouse, such drones can aid a lot in helping to reach the areas which are inaccessible to the workers and check up on the stock. Another example can be that of a real estate company; such companies can benefit from these drones by using them to inspect for cracks, holes or rusts in different types of buildings.
This is why Microsoft and DJI have teamed up to work on AI Drones that can recognize objects, both of these companies have organized an AI Drones Joint University Competition which gathered students from three Hong Kong universities for making the AI models which will help drones in recognizing different types of fruits. The students then released their drones in the air where they had all the fruits scattered around; they then let the drones do their magic. Amazing huh? We think so too.
There were screens which displayed just how successful each of their models were in recognizing the fruits from a scale between lowest to highest, i.e. how a drone was sure that a banana was a banana.
All of these teams carried out this study with the help of using the amazing DJI Phantom 4 Pro which is without any doubt capable of shooting the 4K ultra HD footage, the Microsoft’s AI then scrutinizes it to figure out what it is seeing.
So what exactly is the secret weapon being used here? Well, it is nothing other than edge computing. Putting it more simple and easy to understand it is all about analyzing data which is close to its source (sensors or cameras) which compares to what used to happen, the drone would be able to send the footage to the cloud. Even though computing through the cloud is highly powerful, the transmission time is something that adds a huge lag in the process. Now, this is something which is eliminated by the edge computing by handling the entire analysis on a drone itself.
As soon as a drone can detect something, it will send a message or the whole result back to the respected computer which in turn will be able to make the entire process a lot easier by improving the user experience and the whole performance much better. This is said by Winnie Chu who happens to be a Business Group Lead, Cloud and Enterprise at Microsoft offices in Hong Kong.
However, we are still a million ways away from having these AI drones to be deployed in the real sense, let’s say into the real world out there, for everybody. For starters, even though are many many theoretical uses of such drones, many companies are still in their way of designing the most effective jobs for such drones, they have not yet decided or are fully sure about the useful tasks that they want to assign to such drones. Chu added.
Moreover, for those people who think that these machines might take over their jobs as well, they should not be so fearful since it is not like it’s happening now anyway. Kevin On, the Global Marketing and Communications director of DJI, said that these AI drones are more likely to augment humans instead of replacing them.
He also added that they see drones as empowering workers and that they see them as a new tool which seems to be there in the toolbox, adding to that drones will surely be able to reach those high places which can be quite tricky for humans to reach, in fact, a greater challenge at times. If you are interested in getting Drones for Kids, check our our article on Top 5 Drones for Kids.
Moreover, again if drones are being trained for spotting the fruits, the idea that is being followed here is the ability of these drones to be able to spot on more complex and different shapes in the coming future. These drones will then be able to spot on many complex shapes and this, in the long run, can help save lives just as Andy Kong has pointed out. If someone faints on the mountain and we need something to search and look for that missing person, it would be much easier to use a swarm of drones to look up that person.
As part of this partnership, DJI is also launching a software development kit (SDK) for Windows devices that will allow the Windows developers to make or build native apps for controlling DJI drones. Using the SDK, many developers can also integrate third-party tools for managing payloads or for accessing sensors and robotics components right over their drones. DJI has already offered a Windows-based ground station for it.
“DJI is very excited to form this partnership with Microsoft which is a bit unique, for bringing the power of DJI aerial platforms to the Microsoft developer ecosystem,” Roger Luo, DJI president added. He said that with the help of their new SDK, Windows developers would be able to employ drones soon enough, AI and machine learning technologies for creating intelligent flying robots which will save both businesses as well as time and of course the money. It will also help in making drone technology a mainstay in the workplace.
Interesting because of this pact (Microsoft and DJI have teamed up to work on AI Drones that can recognize objects), Microsoft also stressed that this partnership gives DJI access to its Azure IP Advantage program.
For Microsoft, this partnership is an example of the important role IP plays in making sure that there is a healthy and vibrant technology ecosystem and builds upon the partnerships which exist in emerging sectors such as connected cars as well as the personal wearable,” these were the company’s notes in the announcement.
As amazing and thrilling as it sounds when we talk about how Microsoft and DJI have teamed up to work on AI Drones that can recognize objects, we are hoping that this pact can prove to be extremely useful for both the parties and in our best interests so that our lives could be made a lot easier, which should be the prime concern.