|
|
--- |
|
|
license: apache-2.0 |
|
|
--- |
|
|
|
|
|
Keras Embedding model that generates embeddings for CGI Classification. The model requires that the images be preprocessed by |
|
|
resizing to 256x256 and then converting to grayscale and then applying the Fourier transform. The process can be described as this: |
|
|
1. Resize to 256x256 |
|
|
2. Convert to grayscale |
|
|
3. Apply the Fourier transform |
|
|
4. Feed into model input |
|
|
|
|
|
The model will then transform the input transform into a 128-length embedding vector. These vectors are separable and can be used in classification models. |
|
|
The model doesn't enforce a specific classification model, prototypical networks, XGBoost, and other classifiers can be used to classify. For more detailed instructions refer to [Kaggle](https://www.kaggle.com/models/nhrade/cgi-embedding-detection-module). |
