NCVPRIPG 2023 Challenge

3D Reconstruction and Restoration of Indian Heritage from Partial and Noisy Scans

Problem Definition:

In this challenge, our main focus is to encourage participants to propose novel and efficient classical algorithms to solve 3D Surface Reconstruction. We describe the problem to be solved as follows. Let {Y1,Y2,...,Yk} be a set of partial 3D scans of a site obtained through a 3D scanner. Then, find the 3D model Y that represents the complete 3D model of the underlying 3D surface of the considered site. The reconstructed 3D model Y should be as close as possible to the actual 3D model of the building. For validation of the algorithms, we would provide partial 3D scans and respective ground truth 3D models for some of the architectural sites. For the testing phase, we would only provide partial 3D scans and would ask the participants to submit the reconstructed 3D models.

Dataset:

Please register for getting access to the.

Evaluation: 

In order to rank the submitted algorithms for this challenge we will use the point cloud distance as a ranking metric. In specific, let X be the ground truth point cloud and let Y be the reconstructed point cloud. Then, the Chamfer distance between X and Y will be used as a ranking metric.

Registration:

Please register here if you want to participate in this challenge.

Awards and Recognition

*It is mandatory to attend the NCVPRIPG 2023 in physically in order to be eligible for the awards and certificates.

Rules:

Timeline:

Results:


Organizers:  

Deepanshu Singh Solanki (IIT Jodhpur), Rajendra Nagar (IIT Jodhpur), Prof. Santanu Chaudhury (IIT Jodhpur)

E-mail: rn@iitj.ac.in