2D Classes¶
These are the 2-Dimensional Classes.
Data Representation¶
Data is always represented as an np.ndarray. The dimensions of the array is always 3D.
It has the X, Y and the Z dimension. The Z dimension is referred to as the Channel Dimension.
This is because Z dim holds all channel information!
Despite having a 3D structure, we call helper classes 2D because it’s the source dimension.
Intuition¶
We can think of this representation like a layered cake:
Side View Bird Eye View
(1st Channel)
|-------| (Channel 1) [A][B][C]
|-------| (Channel 2) [B][A][D]
|-------| (Channel 3) [A][B][C]
Each layer is like a channel in the image:
Side View Bird Eye View
(1st Channel)
|-------| (RED) [010][100][075]
|-------| (GREEN) [120][200][100]
|-------| (BLUE) [200][050][000]
[All values are 0-255]
If we were to add more layers, we can just append it as a new layer:
Side View
|-------| (RED)
|-------| (GREEN)
|-------| (BLUE)
|-------| (SPECTRAL_0)
|-------| (SPECTRAL_1)
Hence, the data can be imagined as a 3-dimensional object regardless of how many channels we add.
Channel Dimension¶
The Channel Dimension is where all channel data resides. XY is implicitly represented based on the cell location!
If we were to bring XY information down to the Channel Dimension, they will still be called the XY Channels. This is to keep convention and prevent confusion between Channels, Indices, Spatial Indices etc.
Due to frequency of array indexing, it can become confusing to differentiate array indexes and index values.
There are multiple ways to getting more channels in Frame2D Channels.