pannagas / pointnet-

Created point clouds from epipolar geometry along with PC correspondences using ICP. Later implemented PointNet from scratch to classify and predict the generated point clouds.

Point-cloud-recontruction-and-PointNet-classification

Created point clouds from epipolar geometry along with PC correspondences using ICP. Later implemented PointNet from scratch to classify and predict the generated point clouds.

The first part for creating initial point clouds from multiple views is given in the folder "Sparse reconstruction and point cloud generation". For the second part, PointNet classification and evaluation, I have added a Google Drive link for all the necessary data used for training and evaluation purposes for both - prediction and classification. I have also uploaded my Google Colab link for the entire code. You will have to clone the data on your drive to get the data or change the code to read files directly from the same directory. I have also attached a PDF of some of my outputs for the second part if the results are not visible in the collab link. The results include -

1. Point cloud generation from the depth image
2. Matching 2 Point clouds using ICP
3. Prediction of Point clouds using PointNets
4. Classifying parts of point clouds using PointNets

Link for the Data for PointNet - https://drive.google.com/drive/folders/1lxk3JaunfSKTufIy3DXdC8_iq4QXFRG1?usp=sharing