AI Rank
EN
← 开源
eric-yyjau

pytorch-superpoint

Superpoint Implemented in PyTorch: https://arxiv.org/abs/1712.07629

Model DevelopmentArchitectureJupyter Notebook
在 GitHub 打开
增长势头
+024 小时新增 Star0.0%
934
Star
194
Fork
+0
本周
4
贡献者
创建于 2019-12-14 · 更新于 2026-09-11 · 今日第 12544 名
主要开发者
eric-yyjauguppykangValeyards

eric-yyjau/pytorch-superpoint 排名第 12,544，有 934 个星标。

同类别比较

eric-yyjau/pytorch-superpoint 排名第 12,544，此处有 8 个同类条目；dusty-nv/jetson-reinforcement 排名第 12,546。

同类别比较
名称排名指标
acbull/pyHGT12,529
语言:
Python
星标数:
936
分叉数:
173
简介:
Code for "Heterogeneous Graph Transformer" (WWW'20), which is based on pytorch_geometric
baal-org/baal12,530
语言:
Python
星标数:
936
分叉数:
91
简介:
Bayesian active learning library for research and industrial usecases.
ImprintLab/Medical-SAM212,540
语言:
Python
星标数:
934
分叉数:
139
简介:
Medical SAM 2: Segment 3D Medical Images Via Segment Anything Model 2
benedekrozemberczki/graph2vec12,541
语言:
Python
星标数:
934
分叉数:
165
简介:
A parallel implementation of "graph2vec: Learning Distributed Representations of Graphs" (MLGWorkshop 2017).
dusty-nv/jetson-reinforcement12,546
语言:
C++
星标数:
933
分叉数:
224
简介:
Deep reinforcement learning GPU libraries for NVIDIA Jetson TX1/TX2 with PyTorch, OpenAI Gym, and Gazebo robotics simulator.
rusty1s/pytorch_cluster12,551
语言:
C++
星标数:
933
分叉数:
164
简介:
PyTorch Extension Library of Optimized Graph Cluster Algorithms
Mukosame/Zooming-Slow-Mo-CVPR-202012,566
语言:
Python
星标数:
930
分叉数:
163
简介:
Fast and Accurate One-Stage Space-Time Video Super-Resolution (accepted in CVPR 2020)
microsoft/SoftTeacher12,571
语言:
Python
星标数:
930
分叉数:
123
简介:
Semi-Supervised Learning, Object Detection, ICCV2021
关键参数与规格
关键参数与规格
参数值
语言Jupyter Notebook
类别Model Development
子类别Architecture
星标数934
分叉数194
贡献者4
创建日期2019-12-14
更新日期2026-09-11
1 天星标数0
7 天星标数0
排名12,544
上次排名12,464
排名变化-80
7 天星标排名13,482
上次 7 天星标排名13,490
7 天星标排名变化8
总星标排名11,737
上次总星标排名11,719
总星标排名变化-18
国家USA
README 摘录

pytorch-superpoint This is a PyTorch implementation of "SuperPoint: Self-Supervised Interest Point Detection and Description. " Daniel DeTone, Tomasz Malisiewicz, Andrew Rabinovich. This code is partially based on the tensorflow implementation Please be generous to star this repo if it helps your research. This repo is a bi-product of our paper deepFEPE(IROS 2020). Differences between our implementation and original paper *Descriptor loss*: We tested descriptor loss using different methods, including dense method (as paper but slightly different) and

sparse method. We notice sparse loss can converge more efficiently with similar performance. The default setting here is sparse method. 27 | | superpoint_coco_heat2_0_170k_hpatches_sub | 0. 42 | | superpoint_kitti_heat2_0_50k_hpatches_sub | 0. 41 | Pretrained model is from SuperPointPretrainedNetwork. The evaluation is done under our evaluation scripts. COCO/ KITTI pretrained model is included in this repo. 2) cuda (tested in cuda10) Path setting paths for datasets ($DATA_DIR), logs are set in Dataset Datasets should

be downloaded into $DATA_DIR. The Synthetic Shapes dataset will also be generated there. The folder structure should look like: MS-COCO 2014 HPatches KITTI Odometry run the code Notes: Tensorboard: 1) Training MagicPoint on Synthetic Shapes you don't need to download synthetic data. You will generate it when first running it. 2) Exporting detections on MS-COCO / kitti This is the step of homography adaptation(HA) to export pseudo ground truth for joint training. make sure the pretrained model

in config file is correct make sure COCO dataset is in '$DATA_DIR' (defined in setting. py) config file: General command: export coco - do on training set export coco - do on validation set Edit 'export_folder' to 'val' in 'magicpoint_coco_export. yaml' export kitti config check the 'root' in config file train/ val split_files are included in . 3) Training Superpoint on MS-COCO/ KITTI You need pseudo ground truth labels to traing detectors. Labels can be exported from

step 2) or downloaded from link. Then, as usual, you need to set config file before training. config file root: specify your labels root root_split_txt: where you put the train. txt split files (no need for COCO, needed for KITTI) labels: the exported labels from homography adaptation pretrained: specify the pretrained model (you can train from scratch) 'eval': turn on the evaluation during training General command COCO kitti set your batch size (originally 1) refer to: 'train_tutorial

. md' 4) Export/ Evaluate the metrics on HPatches Use pretrained model or specify your model in config file will run export then evaluation. Export download HPatches dataset (link above). Export keypoints, descriptors, matching evaluate Evaluate homography estimation/ repeatability/ matching scores . 5) Export/ Evaluate repeatability on SIFT Refer to another project: Feature-preserving image denoising with multiresolution filters specify the pretrained model Pretrained models Current best model *COCO dataset* *KITTI dataset* model from magicleap Jupyter notebook Updates (year. 05: Update pytorch

nms from ( Update and test KITTI dataloader and labels on google drive (should be able to fit the KITTI raw

常见问题

记录了多少星标和分叉？

eric-yyjau/pytorch-superpoint 有 934 个星标和 194 个分叉。

仓库排名是多少？

eric-yyjau/pytorch-superpoint 排名第 12,544。

记录的语言和类别是什么？

eric-yyjau/pytorch-superpoint 有 934 个星标，语言为 Jupyter Notebook，类别为 Model Development。

哪个仓库排名最接近？

eric-yyjau/pytorch-superpoint 与 dusty-nv/jetson-reinforcement 比较，后者排名第 12,546，有 933 个星标。

README

pytorch-superpoint

This is a PyTorch implementation of "SuperPoint: Self-Supervised Interest Point Detection and Description." Daniel DeTone, Tomasz Malisiewicz, Andrew Rabinovich. ArXiv 2018. This code is partially based on the tensorflow implementation https://github.com/rpautrat/SuperPoint.

Please be generous to star this repo if it helps your research. This repo is a bi-product of our paper deepFEPE(IROS 2020).

Differences between our implementation and original paper

  • Descriptor loss: We tested descriptor loss using different methods, including dense method (as paper but slightly different) and sparse method. We notice sparse loss can converge more efficiently with similar performance. The default setting here is sparse method.

Results on HPatches

Task Homography estimation Detector metric Descriptor metric
Epsilon = 1 3 5 Repeatability MLE NN mAP Matching Score
Pretrained model 0.44 0.77 0.83 0.606 1.14 0.81 0.55
Sift (subpixel accuracy) 0.63 0.76 0.79 0.51 1.16 0.70 0.27
superpoint_coco_heat2_0_170k_hpatches_sub 0.46 0.75 0.81 0.63 1.07 0.78 0.42
superpoint_kitti_heat2_0_50k_hpatches_sub 0.44 0.71 0.77 0.56 0.95 0.78 0.41
  • Pretrained model is from SuperPointPretrainedNetwork.
  • The evaluation is done under our evaluation scripts.
  • COCO/ KITTI pretrained model is included in this repo.

Installation

Requirements

  • python == 3.6
  • pytorch >= 1.1 (tested in 1.3.1)
  • torchvision >= 0.3.0 (tested in 0.4.2)
  • cuda (tested in cuda10)
conda create --name py36-sp python=3.6
conda activate py36-sp
pip install -r requirements.txt
pip install -r requirements_torch.txt # install pytorch

Path setting

  • paths for datasets ($DATA_DIR), logs are set in setting.py

Dataset

Datasets should be downloaded into $DATA_DIR. The Synthetic Shapes dataset will also be generated there. The folder structure should look like:

datasets/ ($DATA_DIR)
|-- COCO
|   |-- train2014
|   |   |-- file1.jpg
|   |   `-- ...
|   `-- val2014
|       |-- file1.jpg
|       `-- ...
`-- HPatches
|   |-- i_ajuntament
|   `-- ...
`-- synthetic_shapes  # will be automatically created
`-- KITTI (accumulated folders from raw data)
|   |-- 2011_09_26_drive_0020_sync
|   |   |-- image_00/
|   |   `-- ...
|   |-- ...
|   `-- 2011_09_28_drive_0001_sync
|   |   |-- image_00/
|   |   `-- ...
|   |-- ...
|   `-- 2011_09_29_drive_0004_sync
|   |   |-- image_00/
|   |   `-- ...
|   |-- ...
|   `-- 2011_09_30_drive_0016_sync
|   |   |-- image_00/
|   |   `-- ...
|   |-- ...
|   `-- 2011_10_03_drive_0027_sync
|   |   |-- image_00/
|   |   `-- ...

run the code

  • Notes:
    • Start from any steps (1-4) by downloading some intermediate results
    • Training usually takes 8-10 hours on one 'NVIDIA 2080Ti'.
    • Currently Support training on 'COCO' dataset (original paper), 'KITTI' dataset.
  • Tensorboard:
    • log files is saved under 'runs/<\export_task>/...'

tensorboard --logdir=./runs/ [--host | static_ip_address] [--port | 6008]

1) Training MagicPoint on Synthetic Shapes

python train4.py train_base configs/magicpoint_shapes_pair.yaml magicpoint_synth --eval

you don't need to download synthetic data. You will generate it when first running it. Synthetic data is exported in ./datasets. You can change the setting in settings.py.

2) Exporting detections on MS-COCO / kitti

This is the step of homography adaptation(HA) to export pseudo ground truth for joint training.

  • make sure the pretrained model in config file is correct

  • make sure COCO dataset is in '$DATA_DIR' (defined in setting.py)

  • config file:

export_folder: <'train' | 'val'>  # set export for training or validation

General command:

python export.py      [--outputImg | output images for visualization (space inefficient)]

export coco - do on training set

python export.py export_detector_homoAdapt configs/magicpoint_coco_export.yaml magicpoint_synth_homoAdapt_coco

export coco - do on validation set

  • Edit 'export_folder' to 'val' in 'magicpoint_coco_export.yaml'
python export.py export_detector_homoAdapt configs/magicpoint_coco_export.yaml magicpoint_synth_homoAdapt_coco

export kitti

  • config
    • check the 'root' in config file
    • train/ val split_files are included in datasets/kitti_split/.
python export.py export_detector_homoAdapt configs/magicpoint_kitti_export.yaml magicpoint_base_homoAdapt_kitti

3) Training Superpoint on MS-COCO/ KITTI

You need pseudo ground truth labels to traing detectors. Labels can be exported from step 2) or downloaded from link. Then, as usual, you need to set config file before training.

  • config file
    • root: specify your labels root
    • root_split_txt: where you put the train.txt/ val.txt split files (no need for COCO, needed for KITTI)
    • labels: the exported labels from homography adaptation
    • pretrained: specify the pretrained model (you can train from scratch)
  • 'eval': turn on the evaluation during training

General command

python train4.py    --eval

COCO

python train4.py train_joint configs/superpoint_coco_train_heatmap.yaml superpoint_coco --eval --debug

kitti

python train4.py train_joint configs/superpoint_kitti_train_heatmap.yaml superpoint_kitti --eval --debug
  • set your batch size (originally 1)
  • refer to: 'train_tutorial.md'

4) Export/ Evaluate the metrics on HPatches

  • Use pretrained model or specify your model in config file
  • ./run_export.sh will run export then evaluation.

Export

  • download HPatches dataset (link above). Put in the $DATA_DIR. python export.py
  • Export keypoints, descriptors, matching
python export.py export_descriptor  configs/magicpoint_repeatability_heatmap.yaml superpoint_hpatches_test

evaluate

python evaluation.py [-r, --repeatibility | -o, --outputImg | -homo, --homography ]

  • Evaluate homography estimation/ repeatability/ matching scores ...
python evaluation.py logs/superpoint_hpatches_test/predictions --repeatibility --outputImg --homography --plotMatching

5) Export/ Evaluate repeatability on SIFT

# export detection, description, matching
python export_classical.py export_descriptor configs/classical_descriptors.yaml sift_test --correspondence

# evaluate (use 'sift' flag)
python evaluation.py logs/sift_test/predictions --sift --repeatibility --homography
  • specify the pretrained model

Pretrained models

Current best model

  • COCO dataset logs/superpoint_coco_heat2_0/checkpoints/superPointNet_170000_checkpoint.pth.tar
  • KITTI dataset logs/superpoint_kitti_heat2_0/checkpoints/superPointNet_50000_checkpoint.pth.tar

model from magicleap

pretrained/superpoint_v1.pth

Jupyter notebook

# show images saved in the folders
jupyter notebook
notebooks/visualize_hpatches.ipynb

Updates (year.month.day)

Known problems

  • test step 5: evaluate on SIFT
  • Export COCO dataset in low resolution (240x320) instead of high resolution (480x640).
  • Due to step 1 was done long time ago. We are still testing it again along with step 2-4. Please refer to our pretrained model or exported labels. Or let us know how the whole pipeline works.
  • Warnings from tensorboard.

Work in progress

  • Release notebooks with unit testing.
  • Dataset: ApolloScape/ TUM.

Citations

Please cite the original paper.

@inproceedings{detone2018superpoint,
  title={Superpoint: Self-supervised interest point detection and description},
  author={DeTone, Daniel and Malisiewicz, Tomasz and Rabinovich, Andrew},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops},
  pages={224--236},
  year={2018}
}

Please also cite our DeepFEPE paper.

@misc{2020_jau_zhu_deepFEPE,
Author = {You-Yi Jau and Rui Zhu and Hao Su and Manmohan Chandraker},
Title = {Deep Keypoint-Based Camera Pose Estimation with Geometric Constraints},
Year = {2020},
Eprint = {arXiv:2007.15122},
}

Credits

This implementation is developed by You-Yi Jau and Rui Zhu. Please contact You-Yi for any problems. Again the work is based on Tensorflow implementation by Rémi Pautrat and Paul-Edouard Sarlin and official SuperPointPretrainedNetwork. Thanks to Daniel DeTone for help during the implementation.

Posts

What have I learned from the implementation of deep learning paper?

排名相邻
同类项目