Spaces:
Sleeping
Sleeping
Add training notebook and a README
Browse files- README.md +7 -1
- kaggle_training_notebook.ipynb +0 -0
README.md
CHANGED
@@ -10,4 +10,10 @@ pinned: false
|
|
10 |
license: apache-2.0
|
11 |
---
|
12 |
|
13 |
-
|
|
|
|
|
|
|
|
|
|
|
|
10 |
license: apache-2.0
|
11 |
---
|
12 |
|
13 |
+
# Zidane-Scholes-Mbappe Classifier
|
14 |
+
|
15 |
+
A simple image recognition model to classify images as being either Kylian Mbappe, Zinedine Zidane, or Paul Scholes.
|
16 |
+
|
17 |
+
This model is trained using fastai. Using fastai, I got images of each of the players from DuckDuckGo, augmented the data with rotations, skews, crops, and resizes of the images. I then fine-tuned the Resnet34 model to classify images as being one of the three players. I exported the trained model, which I then uploaded to HuggingFace. To see the training process, see kaggle_training_notebook.ipynb
|
18 |
+
|
19 |
+
I initially trained this model by fine-tuning Resnet18, but it made obvious mistakes: it classified Paul Scholes as Mbappe. Moving to Resnet34 makes the test examples work properly
|
kaggle_training_notebook.ipynb
ADDED
The diff for this file is too large to render.
See raw diff
|