Hi Asaf,
Deep learning models like GoogleNet are primarily designed for tasks involving image data, such as classification and segmentation. However, if you want to apply them to your handcrafted features dataset, you can follow some of the below strategies:
- Transform Features: Assess whether the dataset can be transformed into a two-dimensional format, like a matrix or heatmap, and then apply convolutional layers to this transformed data.
- Modify the Model: Adapt the pretrained GoogleNet model to accept non-image data. These models typically expect 2D matrices with three color channels. Since the features are likely 1D vectors, modify the first layer to accept a 1D vector instead of a 3D image. You might also need to replace certain convolutional layers with dense (fully connected) layers that are more suitable for 1D data. Then adjust the final layers to output the desired number of classes.
Hope this will resolve your query!
