Reputation: 61
I'm working with satellite data and I have a DataFrame of pixel values. The satellite image I'm working with is slightly rotated, so I don't have rows and columns with consistent latitude and longitude information (I could back calculate the slope and assign x and y). Can I input this data into a CNN to analyze the structure? I have only seen examples of CNNs where the inputs are arrays of arrays where the structure is retained in a format easily assigned to and x y plane. (I'm new to CNNs)
My data
index Gamma0_VV lat lon dist cell
9355347 0.172325 39.378442 -74.398499 0.003194 0.0
9359296 0.101932 39.378459 -74.398385 0.012222 0.0
9359297 0.116957 39.378549 -74.398407 0.017423 0.0
9359298 0.139411 39.378638 -74.398429 0.025786 0.0
What I've seen
[[0, 0, ... 0, 0,],
[0, 0, ... 0, 0,],
[0, 0, ... 0, 0,]]
Upvotes: 1
Views: 383
Reputation: 121
The sense of CNN is input data to network as matrix. You need to transform this data to matrix and then run CNN network.
You can't use DataFrame as input to CNN
Upvotes: 1