CV 3315 Is All You Need : Semantic Segmentation Competition

TL;DR

This paper reviews transformer-based methods, especially SegFormer, for urban semantic segmentation, achieving a balance of high accuracy and efficiency in a challenging unbalanced dataset competition.

Contribution

It evaluates transformer-driven segmentation models, particularly SegFormer variants, and identifies SegFormer-B2 as the optimal model balancing performance and computational cost.

Findings

SegFormer-B0 achieved 74.6% mIoU with minimal FLOPS.

SegFormer-B5 achieved 80.2% mIoU.

SegFormer-B2 was selected as the final model with 78.5% mIoU and 50.6 GFLOPS.

Abstract

This competition focus on Urban-Sense Segmentation based on the vehicle camera view. Class highly unbalanced Urban-Sense images dataset challenge the existing solutions and further studies. Deep Conventional neural network-based semantic segmentation methods such as encoder-decoder architecture and multi-scale and pyramid-based approaches become flexible solutions applicable to real-world applications. In this competition, we mainly review the literature and conduct experiments on transformer-driven methods especially SegFormer, to achieve an optimal trade-off between performance and efficiency. For example, SegFormer-B0 achieved 74.6% mIoU with the smallest FLOPS, 15.6G, and the largest model, SegFormer- B5 archived 80.2% mIoU. According to multiple factors, including individual case failure analysis, individual class performance, training pressure and efficiency estimation, the final candidate model for the competition is SegFormer- B2 with 50.6 GFLOPS and 78.5% mIoU evaluated on the testing set. Checkout our code implementation at https://vmv.re/cv3315.