one dimension data

[wxw-cyber]

Hello, how do I use tfvan if my data is one dimensional?

[shkarupa-alex]

Hello, in the same way as any other backbone in keras_application.
There are two options:

1. You can repeat your data 3 times across channel dimension and then concat. (I think this is the easiest option)
2. You can transfer pretrained weights to empty model with 1-channel input:

• create 2 models (1 with pretrained weights and 3 channels, 1 with random init and 1 channel)
• find weights in these two models with different shape (first conv weighs)
• sum that weight accross channel dim
• use this summed weight and all other weights from pretrained model and set them to model with 1 weight: model.set_weights([..])

P.S.
Try to search in google how to do that for some backbone like keras resnet50. For tfvan alghoritm will be the same.