OSFormer: One-Stage Camouflaged Instance Segmentation with Transformers

TL;DR

OSFormer introduces a novel one-stage transformer-based framework for camouflaged instance segmentation, effectively combining local and global features to improve accuracy and efficiency with less training data.

Contribution

The paper proposes OSFormer, the first one-stage transformer model for CIS, integrating location-sensing transformers and coarse-to-fine fusion for enhanced performance.

Findings

Achieves 41% AP on camouflaged instance segmentation

Converges efficiently with only 3,040 training samples

Outperforms two-stage frameworks in accuracy and speed

Abstract

We present OSFormer, the first one-stage transformer framework for camouflaged instance segmentation (CIS). OSFormer is based on two key designs. First, we design a location-sensing transformer (LST) to obtain the location label and instance-aware parameters by introducing the location-guided queries and the blend-convolution feedforward network. Second, we develop a coarse-to-fine fusion (CFF) to merge diverse context information from the LST encoder and CNN backbone. Coupling these two components enables OSFormer to efficiently blend local features and long-range context dependencies for predicting camouflaged instances. Compared with two-stage frameworks, our OSFormer reaches 41% AP and achieves good convergence efficiency without requiring enormous training data, i.e., only 3,040 samples under 60 epochs. Code link: https://github.com/PJLallen/OSFormer.