Let the training Begin
My morning routine of getting up early is going well. I got up at 5:30 am, had a shower, and got straight to solving the backpropagation algorithm. Surely, in the mornings my mind reaches a new level of concentration and focus. I got the assignment working in an hour. After which I did my Tacfit commando. Today was the high-intensity day. I believe I pushed myself quite hard. I reached a heart rate of 170 in one of the exercises. After exercising, I felt really good. I once again had a shower as I had a client meeting at 10:30 am. I needed to freshen up and look all clean. I started to collect images to train the neural net for website building. I went through a couple of websites, took screenshots of sections, and stored them in a data training folder.
After I came back, I had lunch and took a short nap. I had to visit the chiropractor as I had a complimentary visit scheduled. It was a rainy day. I took my big umbrella and walked to the chiropractor clinic. It was just a 12 min walk. Close by. She took a few tests and scans of my neck and lower back, made some adjustments, and scheduled another visit to go through the test results. I thought she did a great job. Pleasant, knowledgeable, and welcoming. I did mention my headaches and the frequency at which I get them. Perhaps my lower back issues might be related. You never know.
After I got back, I started to capture more images. I captured around 136 or so section images. I wrote a small script in MATLAB to fetch all the images and convert them to grayscale. After converting the images to grayscale, I had to resize all the images to a fixed resolution. Since my screen capture resulted in capturing images of different resolution, to train a neural net I would need all my images to be the same resolution. By resizing all of them to a fixed resolution I have created an object that will serve as the input to the neural net. I am yet to label all the data. This will be done tomorrow afternoon.
Tomorrow morning I will work on the radio. I will write a driver to send and receive messages from the radio. Once this part is done, I will be able to control the quadcopter remotely.
In the evening, I was drawing out a sketch of what all goes into making a proper quadcopter. What knowledge I need to acquire to have a deeper understanding of making a quadcopter work the way I want it to. I want the quadcopter to be built in-house and vertically integrated. From the motors to the circuitry, to the chassis. For that, I need a good foundation in flight theory, aerodynamics, control theory, attitude and position control, structural and material science, magnetics and motor design, PCB design and EMI issues, and a lot more. Not to mention, light theory, construction of camera, lens, and image processing. There are a lot of branches of science at play to get a quadcopter working. In the beginning, I can be overwhelmed with all the knowledge I need to acquire.
To save me from that pressure, I just need to work on what I already know and slowly build on new knowledge. The goal is to make the quadcopter fly by mid-December. I will need to have the radio working and all the onboard sensors and circuitry like I.M.U, GPS, radio, microcontroller, and drivers all mounted on the quadcopter. I would also need to have a basic mathematical model to create a well-tuned controller.
There is a lot of work to be done. Hard work, consistency, and focus is required. I will need to divide my time well between doing practical work like training a neural net, programming the device drivers, and focusing on acquiring new knowledge, like understanding aerodynamics, new A. I algorithms etc. The ability to balance my energy, focus, and concentration will be the goal for the next 2 weeks.
The courses I will be taking in addition to the Machine learning course are Kinematics: Describing the Motions of Spacecraft by the University of Colorado Boulder and Introduction to Aeronautical Engineering. Both these courses I hope will give me some depth into how I can go about modeling a quadcopter from 1st principle method.
After I came back, I had lunch and took a short nap. I had to visit the chiropractor as I had a complimentary visit scheduled. It was a rainy day. I took my big umbrella and walked to the chiropractor clinic. It was just a 12 min walk. Close by. She took a few tests and scans of my neck and lower back, made some adjustments, and scheduled another visit to go through the test results. I thought she did a great job. Pleasant, knowledgeable, and welcoming. I did mention my headaches and the frequency at which I get them. Perhaps my lower back issues might be related. You never know.
After I got back, I started to capture more images. I captured around 136 or so section images. I wrote a small script in MATLAB to fetch all the images and convert them to grayscale. After converting the images to grayscale, I had to resize all the images to a fixed resolution. Since my screen capture resulted in capturing images of different resolution, to train a neural net I would need all my images to be the same resolution. By resizing all of them to a fixed resolution I have created an object that will serve as the input to the neural net. I am yet to label all the data. This will be done tomorrow afternoon.
Tomorrow morning I will work on the radio. I will write a driver to send and receive messages from the radio. Once this part is done, I will be able to control the quadcopter remotely.
In the evening, I was drawing out a sketch of what all goes into making a proper quadcopter. What knowledge I need to acquire to have a deeper understanding of making a quadcopter work the way I want it to. I want the quadcopter to be built in-house and vertically integrated. From the motors to the circuitry, to the chassis. For that, I need a good foundation in flight theory, aerodynamics, control theory, attitude and position control, structural and material science, magnetics and motor design, PCB design and EMI issues, and a lot more. Not to mention, light theory, construction of camera, lens, and image processing. There are a lot of branches of science at play to get a quadcopter working. In the beginning, I can be overwhelmed with all the knowledge I need to acquire.
To save me from that pressure, I just need to work on what I already know and slowly build on new knowledge. The goal is to make the quadcopter fly by mid-December. I will need to have the radio working and all the onboard sensors and circuitry like I.M.U, GPS, radio, microcontroller, and drivers all mounted on the quadcopter. I would also need to have a basic mathematical model to create a well-tuned controller.
There is a lot of work to be done. Hard work, consistency, and focus is required. I will need to divide my time well between doing practical work like training a neural net, programming the device drivers, and focusing on acquiring new knowledge, like understanding aerodynamics, new A. I algorithms etc. The ability to balance my energy, focus, and concentration will be the goal for the next 2 weeks.
The courses I will be taking in addition to the Machine learning course are Kinematics: Describing the Motions of Spacecraft by the University of Colorado Boulder and Introduction to Aeronautical Engineering. Both these courses I hope will give me some depth into how I can go about modeling a quadcopter from 1st principle method.
The Blog of Anthony Shivakumar
Subscribe to emails from Anthony
First Name
Last Name
Email