bducgroup / deepgps

DeepGPS

This repository is the official PyTorch implementation of DeepGPS, which has been published in IEEE Transactions on Mobile Computing (TMC), a premier journal in the field of mobile computing.

If you find DeepGPS useful, please cite our paper with the following bibtex:

@article{liu2024deepgps,
title={{DeepGPS}: deep learning enhanced {GPS} positioning in urban canyons},
author={Liu, Zhidan and Liu, Jiancong and Xu, Xiaowen and Wu, Kaishun},
journal={IEEE Transactions on Mobile Computing},
volume={23},
number={1},
pages={376--392},
year={2024},
publisher={IEEE}
}

Installation

Requirements: Python >= 3.5, Anaconda3

• Install Pytorch >= 1.8.1

pip install torch==1.8.1+cu111 torchvision==0.9.1+cu111 -f https://download.pytorch.org/whl/torch_stable.html

The latest tested combination is: Python 3.8.5 + Pytorch 1.8.1 + cuda 11.1

Quick Start

python Predict.py

Sample Description

The samples in stored as .npz file. which can be accessed by:

import numpy as np
npz_file = np.load('deepgps.npz')

Each .npz file contains contextual information and ground truth for one position point. The content in npz_file is following:

- npz_file['arr_0']：Environment Matrix.
- npz_file['arr_1']：Four float number. (corrected latitude,corrected longitude, original latitude, original longitude)
- npz_file['arr_2']：The number of seconds that have passed since 00:00:00, 20 June 2019.
- npz_file['arr_3']：The relative coordinate of corrected point (The original point is at (50,50)).
- npz_file['arr_4']：Gaussian Peak Representaiton for ground truth.
- npz_file['arr_5']：Skyplot Matrix.

Download

The weights of a well-trained model can be downloaded from Baidu Cloud Storage (百度网盘) using the following information:

链接：https://pan.baidu.com/s/1ehZHT895ass9PEZ3AsFaTQ?pwd=jb0w
提取码：jb0w

Other Works

Please follow our group at https://github.com/SZU-BDUC/.

deepgps's Issues

DeepGPS model download

Hello:
DeepGPS is a handsome work!!! I would love to download this model to study. But there is an issue when I download this model using 'git lfs clone' or 'git lfs fetch' command. It shows that this repository is over its data quota and account responsible for LFS bandwidth should purchase more data packs to restore access. Meantime, I also fork this repo to my repo, it's also failed to download this model. Hence, I would like to bother you, how can I download this model? Thank you very much. Looking forward to your reply！

您好，模型权重无法下载

权重无法下载，能给个百度网盘的链接吗

作者大大您好，我是今年保研上岸的一枚新生，想咨询几个问题

1.想问问作者大人您是跟随刘志丹老师一起学习吗
2.想问问物联网研究所哪个老师比较好呢，我零基础，本科和嵌入式的一起做过一些东西，但是我还是做软件为主，想稍微接触点硬件的东西

_pickle.UnpicklingError: invalid load key, 'v'.

<\DeepGPS-main> python Predict.py
len: 7
model_path ./trained_model.pkl
Traceback (most recent call last):
File "Predict.py", line 22, in
checkpoint = torch.load(model_path)
File "C:\Users\35381\AppData\Local\Programs\Python\Python38\lib\site-packages\torch\serialization.py", line 593, in load
return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
File "C:\Users\35381\AppData\Local\Programs\Python\Python38\lib\site-packages\torch\serialization.py", line 762, in _legacy_load
magic_number = pickle_module.load(f, **pickle_load_args)
_pickle.UnpicklingError: invalid load key, 'v'.