4Seasons dataset

This pipeline localizes sequences from the 4Seasons dataset and can reproduce our winning submission to the challenge of the ECCV 2020 Workshop on Map-based Localization for Autonomous Driving.

Installation

Download the sequences from the challenge webpage and run:

unzip recording_2020-04-07_10-20-32.zip -d datasets/4Seasons/reference
unzip recording_2020-03-24_17-36-22.zip -d datasets/4Seasons/training
unzip recording_2020-03-03_12-03-23.zip -d datasets/4Seasons/validation
unzip recording_2020-03-24_17-45-31.zip -d datasets/4Seasons/test0
unzip recording_2020-04-23_19-37-00.zip -d datasets/4Seasons/test1

Note that the provided scripts might modify the dataset files by deleting unused images to speed up the feature extraction

Pipeline

The process is presented in our workshop talk, whose recording can be found here.

We first triangulate a 3D model from the given poses of the reference sequence:

python3 -m hloc.pipelines.4Seasons.prepare_reference

We then relocalize a given sequence:

python3 -m hloc.pipelines.4Seasons.localize --sequence [training|validation|test0|test1]

The final submission files can be found in outputs/4Seasons/submission_hloc+superglue/. The script will also evaluate these results if the training or validation sequences are selected.

Results

We evaluate the localization recall at distance thresholds 0.1m, 0.2m, and 0.5m.

Methods	test0	test1
hloc + SuperGlue	91.8 / 97.7 / 99.2	67.3 / 93.5 / 98.7
Baseline SuperGlue	21.2 / 33.9 / 60.0	12.4 / 26.5 / 54.4
Baseline R2D2	21.5 / 33.1 / 53.0	12.3 / 23.7 / 42.0
Baseline D2Net	12.5 / 29.3 / 56.7	7.5 / 21.4 / 47.7
Baseline SuperPoint	15.5 / 27.5 / 47.5	9.0 / 19.4 / 36.4