images might have the roof of the collecting vehicle in it (the camera could also be on a vertical selfie stick held by someone walking, or on the helmet of someone riding a motorcycle, or...)
for our MVP we're just looking at the relative green view index across the dataset, so as long as the camera height above the car is the same and the same vehicle was always used and so the roof is always the same % of the image, we should be fine... but we should ideally have a way to detect if there is a large amount of vehicle in the bottom third of the image, and if there is, then segment it out and mask it in the analysis.
solution should be able to run locally (not dependent on internet/ an API service) and have an open source license.
some examples of images that do have a vehicle roof visible that we want to use include this
images might have the roof of the collecting vehicle in it (the camera could also be on a vertical selfie stick held by someone walking, or on the helmet of someone riding a motorcycle, or...)
for our MVP we're just looking at the relative green view index across the dataset, so as long as the camera height above the car is the same and the same vehicle was always used and so the roof is always the same % of the image, we should be fine... but we should ideally have a way to detect if there is a large amount of vehicle in the bottom third of the image, and if there is, then segment it out and mask it in the analysis.
solution should be able to run locally (not dependent on internet/ an API service) and have an open source license.
some examples of images that do have a vehicle roof visible that we want to use include this
and this
and this
and this