SSNet: Spectral Supplementation Network for Multispectral Pedestrian Detection

Train and Test

• Prerequisites

　 Basic Python(3.8) package installation.

• Run test

python test.py

• Run train

python train.py

Detection Performance

1. kaist dataset

color and thermal

only color

only thermal

2. llvip dataset

only color

only thermal

Remarkably, our method surpasses the CFT method by 0.63% in the mAP metric (mAP50: 97.5, mAP: 63.6), even though the CFT method utilizes both color and thermal modalities simultaneously during the inference phase.