TGRMPT: A Head-Shoulder Aided Multi-Person Tracker and a New Large-Scale Dataset for Tour-Guide Robot
- The tour-guide robot used to collect our dataset and the dataset snapshots are shown as follows.
2. Statistical comparisons between our dataset and existing datasets. K = Thousand, M = Million, min = minutes.
3. The pipeline of our method.
- Download our dataset from Baidu Netdisk. The fetch code is
v0hw
. Assume the dataset is downloaded to ${dataset}
.
- Change working directory to
${dataset}
. Run the following command to extract compressed files.
cd iros2022
find . -type f -name "*.tar" | xargs -i sh -c 'a=`dirname {}`; tar -xf {} -C $a'
- Clone our source code. Assume the code is cloned into
${TGRMPT}
.
- Create local link to dataset.
cd ${TGRMPT} & ln -s ${dataset}/iros2022 dataset
- Detection, see README.
- ReID, see README.
- Tracking and evaluation, see README.
- Deploy, see README.