I have a large 2D LIDAR dataset that I want to feed into a model to classify items in the scan. The 2D data can be represented as a 2D matrix. Each column represents a specific angle from 0-359. Each Row is 1 full sweep around with a new sweep every 25ms. Each data point is the distance at that specific angle at that specific time. The data can also be interpreted as polar data.
Here is an example data slice. I truncated down the columns and rows to useful data and converted it to a greyscale image.
In the center of the image is a feature that I am trying to recognize.
Here is an example with no feature:
I’m looking for some direction on how to create a model for this dataset.
I have had some success converting all the 2D LIDAR data to images and tossing them into cat vs dog from lesson 1. I’m wondering if there are any other suggestions on how to handle this data.
The 2D LIDAR data usually has values ranges from 500 millimeters up to 2500 millimeters and the features I am looking for typically have delta of 20 millimeters.
Has anyone ever dealt with this and have some suggestions?
I have some decent results from cat vs dog with converting the images to greyscale.
Thoughts on trying this as an RNN?