zhyx12/RPA

RPN Prototype Alignment For Domain Adaptive Object Detector

Framework

Example Results

Introduction

In this project, we use Pytorch 1.7.0 and CUDA version is 10.1.

Datasets

Datasets Preparation

• Cityscape and FoggyCityscape: Download the Cityscape dataset, see dataset preparation code in DA-Faster RCNN.
• Sim10k: Download the dataset from this website.

Models

Pre-trained Models

In our experiments, we used two pre-trained models on ImageNet, i.e., VGG16 and ResNet101. Please download these two models from:

• VGG16: Dropbox VT Server
• ResNet101: Dropbox VT Server

Train

CUDA_VISIBLE_DEVICES=$GPU_ID bash ./experiments/scripts/rpn_train.sh train ./configs/rpa/normal_to_foggy.xml

Test

CUDA_VISIBLE_DEVICES=$GPU_ID bash ./experiments/scripts/rpn_train.sh test ./configs/rpa/normal_to_foggy_test.xml