IID: Feature dimension mismatch

[huiyuchen708]

Why is it that when I use the pre-trained Xception model weights for implicit feature extraction and the ResNet34 model weights for explicit feature extraction, the extracted features cannot be used to calculate cosine similarity? The error indicates that the final output dimension of the Xception model is 12048, while the output dimension of the ResNet34 model is 1512.

[huiyuchen708]

Thank you so much! I’ll try it right now.