Class-Wise Fm-Nms For Knowledge Distillation of Object Detection
Lyuzhuang Liu, Tsubasa Hirakawa, Takayoshi Yamashita, Hironobu Fujiyoshi
-
SPS
IEEE Members: $11.00
Non-members: $15.00Length: 00:12:27
This paper deals with automatic 3D reconstruction of objects from frontal RGB images. We propose a complete workflow that can be easily adapted to other object families. First, we detect and segment the object present in the image using Convolutional Neural Networks. in a second step, we perform the final 3D reconstruction of the object by warping the rendered depth maps of a fitted 3D template in 2D image space to match the input silhouette. To explain and validate our method, we focus on 3D guitar reconstructions from real input images and renders of guitar models available in the ShapeNet database. We also show qualitative examples of applying our reconstruction method to other family of objects. The results of this study show that our method can automatically produce high-quality 3D object reconstructions from frontal images using a set of segmentation and 3D reconstruction techniques.