dataset needed for transfer learning

Hi,

I am trying to trained a pre-trained VGG16 classification model for face recognition.
Can you tell me the number of images required for training and validation datasets

Hi Shkoli,
Refer to tlt user guide, tlt provides vgg16 pre-trained model for classification network.
It is available at ngc.nvidia.com.
Customer can also finetune pre-trained model with their own data.
For data, at least 2 classes are needed.But there is no rule for the quantity of each class’ images.

Approximately how many images would be needed to fine tune and train a pre trained model. I have created a dataset that contains 20 images for each of the train/val/test datasets

At ngc.nvidia.com, TLT already provides vgg16 pre-trained model for classification network.
You can use it as a pre-trained model.
For your own dataset, at least 2 classes are needed when train with TLT classification network.But there is no rule for the quantity of each class’ images.

Thanks,
I could train the model using TLT