Dog Vision - ML Algorithm to identify dog breeds
A Machine Learning notebook applied to identify the breed of a dog given an image of a dog.
Problem
Some times, when you are in a bar and a nice doggo passes by, it's nice to know what dog breed is.
Data
The data comes from Kaggle's dog-breed-identification competition.
Evaluation
The evaluation metric used for the competition is Multi Class Log Loss. For this competition submissions have to have a special formatting indicated here.
Features
Some info about the data:
- There are 120 breeds of dogs.
- Around 10k images of dogs in total.
Link to the notebook
If you want to check the full process of transfer learning: here is the full Colab notebook.
On the other hand, if you want to test the algorihm with you custom photo here is the shortened notebook (so you don't need to train a model).
FAQ
You did this by yourself?
Yes and no, I follow up the course of Daniel Bourke (@mrdbourke) on ZTM Academy. So it took a while and I solved a bunch of problems in the process.
Can I build it myself?
Of course, you need to download the data from Kaggle (~700 Mb), connect your Google Drive and do it!.
I have another question, where can I find you?
You have different options:
- Submit an Issue,
- Send me an email at: aucacoyan@gmail.com, or
- Find me on Discord AucaCoyan#9411