You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am currently new to all this and still learning. Is your implementation very similar to FaceNet approach? As I am currently working on a facial recognition project and I am trying FaceNet approach. I have read David Sandberg's on FaceNet and noticed he used a model architecture of inception ResNet v1 to train the dataset.
May I know what is the model architecture that you are using and if I were to use your approach, must I retrain a huge dataset of photos and use this pretrained model, to train my own much smaller dataset? Any suggestions of what I need to do? Thank you.
The text was updated successfully, but these errors were encountered:
Since my example works with MNIST, I'm not using a very big architecture (just 2 convolutions + fully connected).
For face recognition, you need a bigger model like resnet. If your dataset is small, you may want to start with a pretrained model and fine-tune it, or even freeze most of the layers.
Hi,
I am currently new to all this and still learning. Is your implementation very similar to FaceNet approach? As I am currently working on a facial recognition project and I am trying FaceNet approach. I have read David Sandberg's on FaceNet and noticed he used a model architecture of inception ResNet v1 to train the dataset.
May I know what is the model architecture that you are using and if I were to use your approach, must I retrain a huge dataset of photos and use this pretrained model, to train my own much smaller dataset? Any suggestions of what I need to do? Thank you.
The text was updated successfully, but these errors were encountered: