Week #3 — MobileNet

Sachin C -

This week, I took a break from data collection to learn more about MobileNet, the architecture I intend to use to develop my model.

What is MobileNet?

MobileNet is a convolutional neural network (CNN) optimized for devices with limited computing power. This makes it a good candidate for microcontrollers. It was introduced by Google in 2017.

How Does MobileNet Work?

Unlike traditional CNNs,  MobileNet uses depthwise separable convolutions. This splits convolutions into two bilateral processes:

  1. Depthwise Convolution: Applies a single convolutional filter to each input channel separately, reducing the computational cost,
  2. Pointwise Convolution: Uses a 1×1 convolution to combine the outputs of the depthwise convolution, reducing the number of parameters.

Essentially, instead of processing all the data individually, it splits it into multiple processes and combines the results at the end. This ultimately contributes to reducing the amount of energy a model consumes while it processes data.

Next week, I will likely continue finding data sources. I will also consider using tools such as Python to automate this process, although I have made significant process on the data myself

 

Thank you for reading, and I’m excited to see you next week!

More Posts

Comments:

All viewpoints are welcome but profane, threatening, disrespectful, or harassing comments will not be tolerated and are subject to moderation up to, and including, full deletion.

    shreyash_p
    Great job Sachin! I really like how you explained MobileNet’s depthwise separable convolutions and why they’re so important for running on low-power devices. Your plan to continue finding data sources and possibly automate the process with Python also sounds very promising. Keep it up!
    March 1, 2025 at 10:25 pm - Reply
    camille_bennett
    Hi Sachin, sounds like great work. Can you clarify what a convolution is?
    March 4, 2025 at 12:32 pm - Reply
    sachin_c
    Ms. Bennett, A convolution is a complex mathematical operation that allows an AI model to extract certain features of an image (colors, shapes, textures, images, etc) in order to identify it more efficiently. In the context of my project, DCs and PCs (depthwise convolutions and pointwise convolutions) are methods where we use various input channels (the way a model receives data) to identify multiple features at once, and then combine the outputs of those input channels to reduce the number of parameters, or the 'inputs' for the AI model.
    March 5, 2025 at 11:36 am - Reply

Leave a Reply

Your email address will not be published. Required fields are marked *