HFNET Application: Locate yourself in the mall just by taking a photo!

Run application.ipynb .
商场内拍照定位应用。任意视角拍摄一张商户照片，与实现数据库照片比对，返回数据库对应照片的坐标，解决室内GPS较弱情况下的定位问题。 Reference: https://github.com/ethz-asl/hfnet

train_explore: Process of exporting labels of google landmarks with superpoint and netVLAD, and then distill to get the student model.

The outputs are saved in EXPER_PATH, which is too large to be uploaded here but is available on Google Drive.

output path: hfnet/dataset/EXPER_PATH

test_explore: Process of evaluating the trained models.

The datasets are downloaded as indicated in the dataset documentation. SfM models of Aachen, RobotCar, CMU, and Extended CMU, built SuperPoint and usable with HF-Net, are provided here. Download and unpack the HF-Net weights in hfnet/dataset/EXPER_PATH/hfnet/. To localize with NV+SP, download the network weights of NetVLAD and SuperPoint and put them in hfnet/dataset/DATA_PATH/weights/.

Superpoint

Reference: https://github.com/rpautrat/SuperPoint

Process of training with synthetic dataset: Superpoint_train_explore.ipynb

output : https://drive.google.com/drive/folders/1tHNplSRgd6IUNkYT9LgR3UAM6HTM-7D1?usp=sharing

output path: SuperPoint/SuperPointDATA/EXPER_DIR

MS-COCO 2014 and HPatches should be downloaded into SuperPoint/SuperPointDATA/DATA_DIR. The Synthetic Shapes dataset will also be generated there. The folder structure should look like:

$DATA_DIR
|-- COCO
|   |-- train2014
|   |   |-- file1.jpg
|   |   `-- ...
|   `-- val2014
|       |-- file1.jpg
|       `-- ...
`-- HPatches
|   |-- i_ajuntament
|   `-- ...
`-- synthetic_shapes  # will be automatically created