| dc.description.abstract | A Bidirectional Reflectance Distribution Function (BRDF) characterizes the reflectance
behavior of materials to define their optical properties. Bidirectional reflectance
can be used to understand the visual appearance properties of the material.
However, these measurements can be dense in nature and huge in volume and,
therefore, can be challenging to record and store. To resolve this, several reflectance
models are proposed for the estimation of the bidirectional reflectance properties
of the material.
The traditional BRDF estimation process is time-consuming and expensive. To
make this estimation faster and more efficient, the implementation of machine
learning to estimate BRDF measurements for different materials is introduced
in this thesis. This implementation will be challenging yet effective because of
the nature of the BRDF measurements data. The research is divided into two
parts- classifying the measurements to find what materials they belong to and
implementing regressor models to estimate BRDF measurements of materials at
the specific incident and outgoing directions. The whole research work contains
measuring of the reflectance properties of different sample materials in the lab,
training machine learning models by using those measured data, and estimating
BRDF measurements using machine learning models. | |