ResNet50/FPN Semantic Segmentation

Driven by a personal expedition to advance object detection capabilities, I delved into semantic segmentation models. I developed a robust understanding of their principles before training and testing a model on a diverse dataset. This project utilized a unique encoder-decoder architecture, combining ResNet-50 as the encoder with FPN as the decoder. I utilized PyTorch and Tensorflow framework variables and parameters to train the model, utilizing CUDA runtime and NVIDIA GPU drivers. The resulting model provides a detailed understanding of drivable space and enables precise classification of surrounding environmental variables.