Forests are the nature’s carbon sinks and from last few decades the carbon footprint in our environment has increased exponentially due to heavy industrialisation whereas the forests are being depleted and used as a biofuel. So, there is a need to account for the forest biomass in our environment in order to have the clear statistics about how much carbon emissions can these forests reduce and how much reforestation should be done to tackle the increasing carbon footprint.

The above-mentioned issue can be addressed using a LiDAR based machine learning approach which can give us the biomass estimation. However, the afore-mentioned solution is quite expensive due to the equipment cost, huge time consumption and needs high technical skills to operate. The idea here is to work with monocular vision rather than using LiDAR. The system would work by using semantic segmentation, segmenting the canopy of trees and taking DBH (diameter at breast height) of trees as a parameter and ultimately resulting the biomass content of the tree. The system would further be improved by testing other state of the art deep learning models to reduce the inference time.