Problem Statement Link to heading
As of today, I have a three-year-old dog that suffers from separation anxiety. We have tried various techniques to reduce the anxiety but have had no luck. From trainers, toys, and exercise, we faced the exact outcome of being unable to leave my dog for more than 10 minutes without her howling and pacing.
Motivation Link to heading
I want to combine what I have been doing to solve this issue with a more data-centric approach. I would like to see what works and what does work for my dog. Instead of applying a one-size-fits-all solution, let’s get more personalized. In a nutshell, I will be capturing videos and processing the videos to collect the following metrics;
- Date and time of training
- Pacing between two points
- Number of howls
Diagram Link to heading
- Run a script to capture video from my laptop
- Run a script to process the video to reduce the size
- Run a script that uses a hugging face model to generate predictions
- Run a script to see if dog predictions are pacing or not pacing
- Run a script to process audio for dog howls
- Compress predictions, metrics, and video to s3
- Upload tar gzip file to s3
Code Link to heading
Currently in private repo and will make public soon.
Learnings Link to heading
Here are the things I answered for myself while prepping this post and doing this project.
- How do I use hugo shortcodes ?
- How can I leverage open-source deep learning models for detection?
- How do I upload to s3?
- How do I capture and process video and audio with Python?
A cute clip that will make your day! Rick Roll 🏋🏾♂️