Technical Report of 2023 ABO Fine-grained Semantic Segmentation Competition

TL;DR

This report details the technical approach and experimental findings of a semantic segmentation model for 3D product shapes, achieving third place in a competitive challenge through optimized training strategies.

Contribution

It introduces a fine-grained 3D semantic segmentation method using DGCNN with specific training techniques, advancing the accuracy in classifying product shape categories.

Findings

Learning rate warm restarts improve model performance.

Category-specific learning rate adjustments are beneficial.

Achieved 3rd place in the ICCV 3DVeComm Challenge.

Abstract

In this report, we describe the technical details of our submission to the 2023 ABO Fine-grained Semantic Segmentation Competition, by Team "Zeyu\_Dong" (username:ZeyuDong). The task is to predicate the semantic labels for the convex shape of five categories, which consist of high-quality, standardized 3D models of real products available for purchase online. By using DGCNN as the backbone to classify different structures of five classes, We carried out numerous experiments and found learning rate stochastic gradient descent with warm restarts and setting different rate of factors for various categories contribute most to the performance of the model. The appropriate method helps us rank 3rd place in the Dev phase of the 2023 ICCV 3DVeComm Workshop Challenge.