The 3rd SHApe Recovery from Partial textured 3D scans (SHARP) Workshop and Challenge will be held in conjunction with CVPR on June 19, 2022. The goal of this competition (Challenge 2 - Track 2) is to promote the development of methods to recover fine details from raw scans. Given a 3D object scan with smooth edges, the goal of this challenge is to reconstruct the corresponding CAD model as a triangular mesh with sharp edges approximating the ground-truth sharp edges.
Given a corrupted 3D object scan, X
, the goal is to recover sharp edges E
and the geometry of the corresponding CAD model as a triangular mesh, Y
. The corrupted 3D object scan, X
, were obtained by virtually scanning the CAD models, Y
, using a proprietary 3D scanning pipeline developed by Artec3D. The pairs of CAD models and 3D scans are pre-aligned. Note, that 3D scans may have artifacts in the form of missing parts and protrusions due to specifics of a scanning system. The sharp edges, E
, extracted from the CAD models, Y
, are given as a collection of parametric curves among three types: circles, lines, and splines.
Routines to load and sampmle points on parametric edges are provided. These routines and further documentation can be found in the GitLab repository of SHARP 2022.
The scan shape recovery:
The Sharp Edge recovery:
The quality of estimated sharp edges are evaluated quantitatively w.r.t the ground-truth sharp edges. The evaluation of the edges is conducted using two criterion:
Consist of two directed distances:
The directed distance between edges A and B is approximated in practice by sampling points on A and computing their distances to the nearest lines in edges B. If a parametric sharp edge (PSE) is esimated as a set of coarse Line Sets, points are uniformly sampled w.r.t to the length of the edges.
For each sampled point in A, the Euclidean distance to the nearest line is computed using a point-to-line distance calculation over all the edges of B. The distances are then averaged for all the points. This bi-directional ditance erros are computed for all the PSE in the predictions and ground truth respectively.
Consists of a score that quantifies the similarity between the edge length of the estimation and that of the reference. The length of the estimated edges and the reference edges are computed by summing over the lengths of the edges of each edge-set.
The edge length score is obtained using a similar strategy to surface area scores and mapped into a value in [0,1]. Good estimations are expected to have high length scores.
The total Edge Recovery Score is computed by summing over edge-to-edge distance scores, mapping them into a score in [0,1] and multiplying the result by the edge length score.
The final score is computed as weighted sum of Edge Recovery Score and Shape Reconstruction Score.
More details can be found in the evalation documentation of GitLab repository of SHARP 2022.
If you are interested in joining the challenge, please register https://cvi2.uni.lu/sharp2022/registration.
Please refer to the documents sent by email following your registration on https://cvi2.uni.lu/sharp2022/registration.
The data is to be shared once the data license agreements are signed by all parties. The data license agreements are shared with the participants following their registrations on https://cvi2.uni.lu/sharp2022/registration/.
Start: May 2, 2022, midnight
Description: In this evaluation phase, the participants are asked to submit their final predictions. The scoring program will be automatically launched after submitting the predictions following the format decribed in the submission format page. Multiple submissions are allowed.
May 24, 2022, 11:59 a.m.
You must be logged in to participate in competitions.
Sign In