Determine number of input neurons in image segmentation neural network using fastai?

user_13529782 0
Viewed 134

Hello I am new to neural networks and specifically layers and neurons. I am doing image segmentation project were I have 200 input images and have labeled correctly. In fastai you are able to use a pre-trained model for transfer learning.I have finished training, used learn.summarry. I am confused on how I can determine the number of neurons of in the input layer( initial data)? I've believe it has something to with the shape but I am not sure how to find out the number of input neurons in input layer. May someone help me please! Thank You!


ImageDataBunch;

Train: LabelList (160 items)
x: SegmentationItemList
Image (3, 128, 128),Image (3, 128, 128),Image (3, 128, 128),Image (3, 128, 128),Image (3, 128, 128)
y: SegmentationLabelList
ImageSegment (1, 128, 128),ImageSegment (1, 128, 128),ImageSegment (1, 128, 128),ImageSegment (1, 128, 128),ImageSegment (1, 128, 128)
Path: /content/drive/My Drive/Umes2020/JPEGImages;

Valid: LabelList (40 items)
x: SegmentationItemList
Image (3, 128, 128),Image (3, 128, 128),Image (3, 128, 128),Image (3, 128, 128),Image (3, 128, 128)
y: SegmentationLabelList
ImageSegment (1, 128, 128),ImageSegment (1, 128, 128),ImageSegment (1, 128, 128),ImageSegment (1, 128, 128),ImageSegment (1, 128, 128)
Path: /content/drive/My Drive/Umes2020/JPEGImages;
image-segmentation fast-ai deep-learning neural-network python
1 Answers

I am assuming you meant the number of channels in the input since the number of neurons of the first layer is not limited by the image shape.

For a regular CNN, the number of channels is 3 (RGB). For a fully connected network, the number of channels is 3*H*W since every pixel in every color channel (RGB) becomes 1 flattened channel.

user_32374380
Related