{"id":13443642,"url":"https://github.com/jialeli1/From-Voxel-to-Point","last_synced_at":"2025-03-20T17:30:45.636Z","repository":{"id":217418785,"uuid":"383694948","full_name":"jialeli1/From-Voxel-to-Point","owner":"jialeli1","description":"\"From Voxel to Point: IoU-guided 3D Object Detection for Point Cloud with Voxel-to-Point Decoder\" and \"Anchor-free 3D Single Stage Detector with Mask-Guided Attention for Point Cloud\" in ACM MM 2021.","archived":false,"fork":false,"pushed_at":"2021-11-24T08:03:24.000Z","size":333,"stargazers_count":40,"open_issues_count":0,"forks_count":7,"subscribers_count":4,"default_branch":"main","last_synced_at":"2024-10-28T07:39:32.835Z","etag":null,"topics":["3d-object-detection","pointcloud"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/jialeli1.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null}},"created_at":"2021-07-07T06:21:27.000Z","updated_at":"2024-08-21T12:29:49.000Z","dependencies_parsed_at":"2024-01-16T09:22:13.123Z","dependency_job_id":null,"html_url":"https://github.com/jialeli1/From-Voxel-to-Point","commit_stats":null,"previous_names":["jialeli1/from-voxel-to-point"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jialeli1%2FFrom-Voxel-to-Point","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jialeli1%2FFrom-Voxel-to-Point/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jialeli1%2FFrom-Voxel-to-Point/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/jialeli1%2FFrom-Voxel-to-Point/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/jialeli1","download_url":"https://codeload.github.com/jialeli1/From-Voxel-to-Point/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":244660156,"owners_count":20489296,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["3d-object-detection","pointcloud"],"created_at":"2024-07-31T03:02:06.016Z","updated_at":"2025-03-20T17:30:45.024Z","avatar_url":"https://github.com/jialeli1.png","language":"Python","funding_links":[],"categories":["Python"],"sub_categories":[],"readme":"# FromVoxelToPoint \u0026 MGAF-3DSSD\n\nThis is a reproduced repo of \"From Voxel to Point: IoU-guided 3D Object Detection for Point Cloud with Voxel-to-Point Decoder\" ([FromVoxelToPoint](https://arxiv.org/pdf/2108.03648.pdf)) and \"Anchor-free 3D Single Stage Detector with Mask-Guided Attention for Point Cloud\" ([MGAF-3DSSD](https://arxiv.org/pdf/2108.03634.pdf)) in ACM MM 2021. \n\nThe code is mainly based on [OpenPCDet](https://github.com/open-mmlab/OpenPCDet).\n\n## Introduction\nWe provide codes and training configurations of FromVoxelToPoint \u0026 MGAF-3DSSD on the KITTI and Waymo datasets. Checkpoints will not be released.  \n\n\u003c!-- **Important Notes**: VoTr generally requires quite a long time (more than 60 epochs on Waymo) to converge, and a large GPU memory (32Gb) is needed for reproduction. --\u003e\n\u003c!-- Please strictly follow the instructions and train with sufficient number of epochs. --\u003e\n\u003c!-- If you don't have a 32G GPU, you can decrease the attention SIZE parameters in yaml files, but this may possibly harm the performance.  --\u003e\n\n## Requirements\nThe codes are tested in the following environment:\n* Ubuntu 20.04.1 LTS \n* Python 3.6\n* PyTorch 1.7.1+cu110\n* CUDA 11.0\n* OpenPCDet v0.3.0 (You can easily add the relevant codes to the latest OpenPCDet if you want.)\n\u003c!-- * spconv v1.2.1 --\u003e\nNote that we use a modified spconv to avoid sudo permission requirements during the installation process, which can be easily installed by setup.py.\n\n\n## Installation\na. Clone this repository.\n```shell\ngit clone https://github.com/jialeli1/From-Voxel-to-Point.git\n```\n\nb. Install the dependent python libraries as follows:\n```shell\npip install -r requirements.txt \n```\n\nc. Compile CUDA operators by running the following command:\n* CUDA ops in OpenPCDet and the useful spconv.\n```shell\npython setup.py develop\n```\n* Deformable convolution that is modified from [Deformable-Convolution-V2-PyTorch](https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch.git).\n```shell\ncd pcdet/ops/DeformableConvolutionV2PyTorch\nsh make.sh\n```\n\n## Dataset Preparation\nWe provide model configurations on KITTI and Waymo. Please follow OpenPCDet to prepare the datasets. You can also use \"ln -s\" to link an existing dataset here for a quick start.\n\n## Training\nDetails are in paper. If you use different number of GPUs for training, it's necessary to change the respective training epochs to attain a decent performance. \n\nYou can run training and evaluation commands following OpenPCDet. We also provide some examples on KITTI as follows.\n\n### KITTI\n* models\n```shell\n# MGAF-3DSSD: An RTX 3090 GPU (24G) can contrain 4 KITTI point clouds for training.\ntools/cfgs/kitti_models/MGAF-3DSSD/mgaf-3dssd.yaml\ntools/cfgs/kitti_models/MGAF-3DSSD/mgaf-3dssd_3classes.yaml\n\n# FromVoxelToPoint: An RTX 3090 GPU (24G) can contrain 3 KITTI point clouds for training. It requires a large GPU memory for reproduction.\ntools/cfgs/kitti_models/FV2P/fv2p.yaml\ntools/cfgs/kitti_models/FV2P/fv2p_3classes.yaml\n```\n\n* training on KITTI\n```shell script\ncd tools\n\nCUDA_VISIBLE_DEVICES=6,7 bash scripts/dist_train.sh 2 --cfg_file ./cfgs/kitti_models/MGAF-3DSSD/mgaf-3dssd.yaml\n\nCUDA_VISIBLE_DEVICES=4,5,6,7 bash scripts/dist_train.sh 4 --cfg_file ./cfgs/kitti_models/FV2P/kitti_fv2p.yaml\n```\n\n* evaluation on KITTI\n```shell\ncd tools\n\nCUDA_VISIBLE_DEVICES=7 python test.py --cfg_file ./cfgs/kitti_models/MGAF-3DSSD/mgaf-3dssd.yaml --eval_all\n```\n\n### Waymo\n* models\n```shell\n# MGAF-3DSSD: \ntools/cfgs/waymo_models/MGAF-3DSSD/waymo_mgaf-3dssd_e36.yaml\n\n# FromVoxelToPoint: \ntools/cfgs/waymo_models/FV2P/waymo_fv2p_e30.yaml\n```\n\n\n## Citation \nIf you find this project useful in your research, please consider cite:\n\n```\n@inproceedings{fv2p_mm21,\n  author    = {Jiale Li and\n               Hang Dai and\n               Ling Shao and\n               Yong Ding},\n  title     = {From Voxel to Point: IoU-guided 3D Object Detection for Point Cloud\n               with Voxel-to-Point Decoder},\n  booktitle = {{MM} '21: {ACM} Multimedia Conference},\n  pages     = {4622--4631},\n  year      = {2021},\n}\n\n@inproceedings{mgaf_mm21,\n  author    = {Jiale Li and\n               Hang Dai and\n               Ling Shao and\n               Yong Ding},\n  title     = {Anchor-free 3D Single Stage Detector with Mask-Guided Attention for\n               Point Cloud},\n  booktitle = {{MM} '21: {ACM} Multimedia Conference},\n  pages     = {553--562},\n  year      = {2021},\n}\n\n```","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjialeli1%2FFrom-Voxel-to-Point","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fjialeli1%2FFrom-Voxel-to-Point","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjialeli1%2FFrom-Voxel-to-Point/lists"}