In this project, the PointNet model is implemented for 3D Point Cloud Segmentation using the SemanticKITTI dataset in PyTorch. Out of the available 1000 samples, only 750 were utilized for this particular endeavor. The PointNetLoss function was applied during training, leading to an achieved accuracy of 82% after 100 epochs. The focus of the project was to showcase the model's performance on a subset of the dataset, providing valuable insights into its segmentation capabilities. The utilization of PyTorch facilitated efficient implementation and experimentation throughout the training process. The model was trained on a GTX 3060.
Dataset: Link
To delve into RGBD images, please refer to the following link: Link.