Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

exact working steps for conda env #3

Open
wants to merge 106 commits into
base: master
Choose a base branch
from
Open
Changes from 1 commit
Commits
Show all changes
106 commits
Select commit Hold shift + click to select a range
24b47d4
Fixed misc issues to work with spconv2
Mar 30, 2023
47a5a05
removed old splits
Mar 30, 2023
3ff94e4
Fixed distributed training single machine and added coda dataloader
Apr 1, 2023
688e24a
Updated dockerfile for higher cuda version
Apr 1, 2023
a1b6527
Updated docker file and pvrcnn
Apr 2, 2023
f73eb5a
Updated waymo data gen issue
Apr 2, 2023
7e0fdac
Updated dist training files, coda training, and oracle
Apr 3, 2023
3732083
Added directory check before processing sample list
Apr 3, 2023
faf3d29
Updated launcher files
Apr 3, 2023
5da32d2
Added nuscenes to coda da
Apr 3, 2023
40c7618
Fixed launchers
Apr 3, 2023
92a6ecb
Updated shift coord for coda
Apr 3, 2023
46512c5
Added coda class balanced sampling
Apr 4, 2023
60f2e3a
Added eval script for nuscenes coda adaptation
Apr 4, 2023
7a408ee
Updated batch size and eval
Apr 4, 2023
6883a51
Fixed slurn test script typo
Apr 4, 2023
b63ffca
Added nuscenes mods for debugging
Apr 5, 2023
510e5da
Added fixed up nuscenes train files
Apr 5, 2023
6003f12
Updated gitignore and train slurn scripts
Apr 5, 2023
6a25a01
Updated batch size for training coda
artzha Apr 6, 2023
8e40828
Updated dist train models
Apr 6, 2023
69c613f
Merge branch 'master' of github.com:artzha/ST3D
Apr 6, 2023
c29e8ea
Added self train utils for nuscenes to coda
artzha Apr 6, 2023
ef3c0c2
Merge branch 'master' of github.com:artzha/ST2D
Apr 6, 2023
a12186c
Updated dist test scripts
artzha Apr 6, 2023
a5c069f
Added pvrcnn training utils
artzha Apr 6, 2023
a4e9068
Added training files for waymo
Apr 7, 2023
643c775
Merge branch 'master' of github.com:artzha/ST3D
Apr 7, 2023
7cc0e23
Updated stat norm for waymo coda da
Apr 7, 2023
525c0b8
Major fixes to st3d multiprocessing, batched waymo eval and finetuning
Apr 12, 2023
2d0acf7
Added ls script changes
Apr 12, 2023
cdaccc4
Merge branch 'st3d++' of github.com:artzha/ST3D into st3d++
Apr 12, 2023
bad0654
Tuned pvrcnn st3d params to work better for coda
Apr 13, 2023
d79445b
Added finetuning head configs
Apr 22, 2023
262b78a
Added wandb logging and hyperparameter tuning
Apr 23, 2023
fe95016
Merge branch 'st3d++' of github.com:artzha/ST3D into st3d++
Apr 24, 2023
a440f35
Added items for launching wandb training
Apr 28, 2023
2b5329c
Adapted Centerpoint Pillar Backbone to ST3D
Apr 30, 2023
a7a1c1b
Ignoring wandb files
Apr 30, 2023
72e8d91
Fixed centerpoint eval issue
May 5, 2023
a9b386d
Enabled waymo and nuscenes eval logging on wandb
May 6, 2023
4dd8586
Merge branch 'st3d++' of github.com:artzha/ST3D into st3d++
May 8, 2023
8186ff4
Added training files for pointpillars for nuscenes and waymo
May 8, 2023
8a00f17
Added pp nuscenes training
May 9, 2023
6dcc77b
Refactored iou partly to handle large datasets and single batch
May 11, 2023
6cd818b
Merge branch 'st3d++' of github.com:artzha/ST3D into st3d++
May 11, 2023
b023216
Added trainable versions of all models
May 19, 2023
9884ecd
Merge branch 'st3d++' of github.com:artzha/ST3D into st3d++
May 19, 2023
8734c62
Added kitti train files
May 19, 2023
efae3d1
Added prediction saver
May 22, 2023
73ff7b5
Added training files for coda multi class
May 23, 2023
f4cf560
Added training files for coda to waymo nuscenes
May 27, 2023
8a9cafe
Added modifications to train files
May 28, 2023
cac5e27
Added pretrain to coda files
May 29, 2023
3540ba9
Added cods 32/128 to nuscenes
May 29, 2023
ca089ed
Merge branch 'st3d++' of github.com:artzha/ST3D into st3d++
May 29, 2023
b52d9a1
Decreased learning rate for da models
May 29, 2023
cf96fe2
Merge branch 'st3d++' of github.com:artzha/ST3D into st3d++
May 29, 2023
7a31885
Added pretraining files
May 30, 2023
1eaa54e
Added finetuning lr for coda coda
May 30, 2023
e9ba86c
Merge branch 'st3d++' of github.com:artzha/ST3D into st3d++
May 30, 2023
2c68d26
Added automatic multidataset evaluation to test.py
May 30, 2023
e0c051d
Merge branch 'st3d++' of github.com:artzha/ST3D into st3d++
May 30, 2023
5f50f18
moved evaluation to coda oracle directory
May 30, 2023
325aadb
Added eval configs for lower res lidar channels
May 30, 2023
a26e0f5
Fixed multi config evaluation in test.py
May 31, 2023
412aa48
minor changes to fix naming on models
Jun 20, 2023
5a25c7a
Added training files for coda resolutions and slurn launcher
Jun 20, 2023
4bf6dad
Merge branch 'st3d++' of github.com:artzha/ST3D into st3d++
Jun 20, 2023
e1b5f5c
Added openpcdet training files for kitti pointpillars
Jun 21, 2023
5b41c9c
Added nuscenes openpcdet cfgs
Jun 21, 2023
c5f1e9c
Merge branch 'st3d++' of github.com:artzha/ST3D into st3d++
Jun 21, 2023
331d2ca
Added working centerpoint voxel backbone for training
Jun 21, 2023
66963d7
Added several changes to model parameters to reflect openpcdet model …
Jun 21, 2023
3f64ed7
Cleaned up train files and added the lonestar job script
Jun 21, 2023
dc8c5ff
Fixed evaluation for waymo on 20% of test and added sota model configs
Jun 25, 2023
4f73b5e
Updated nuscenes to coda training configs
Jun 27, 2023
7fba274
Fixed path lookup issue for tests and added config files
Jun 27, 2023
5df151b
Added waymo specific coda training files
Jun 28, 2023
aef64e8
Added coda to waymo cfgs
Jun 28, 2023
0f32562
Added new model configs for coda to av transfer
Jun 29, 2023
3578487
Updated st3d to pcdet 0.5 and added model configs for st3d training
Jun 29, 2023
a4dd971
Update README.md
artzha Jun 30, 2023
0b4ac7f
Added finetuning head configs for new pvrcnn config
Jul 1, 2023
37297b1
Merge branch 'st3d++' of github.com:artzha/ST3D into st3d++
Jul 1, 2023
0a62a11
Added coda coda finetune heafull config files and fix to 1 dim predic…
Jul 3, 2023
d1f529f
Modified how dist train evaluates datasets and added single stage gt …
Jul 3, 2023
764b8e4
Added single stage waymo coda finetuning script
Jul 4, 2023
20c6367
Added nuscenes pretraining for coda cfg
Jul 4, 2023
1386bd4
Fixed naming waymo single stage config issue
Jul 4, 2023
de10e9e
fixed nuscenes training config lanucher
Jul 4, 2023
e740a5c
Added allclass coda to av training
Jul 5, 2023
0b1144c
Added train files for coda 32 finetune headfull
Jul 8, 2023
c960a2b
Made evaluation of train dataset before target datasets
Jul 8, 2023
cfc1ad0
Merge branch 'st3d++' of github.com:artzha/ST3D into st3d++
Jul 8, 2023
53d8903
Added coda128 to waymo train config file
Jul 9, 2023
95e0cae
Reduced batch size for coda128 to waymo training to fit on gpu
Jul 9, 2023
3263bed
Added training config for coda128 to nus
Jul 9, 2023
1e86a53
Turned on wandb logging for coda to nus
Jul 9, 2023
f1d39db
Updated training files to include st3d pretraining
Jul 11, 2023
9d58a04
Modified training configs for st3d
Jul 11, 2023
d455f31
Added model training configs and readme update
Jul 12, 2023
8144316
Merge pull request #1 from artzha/st3d++
artzha Jul 12, 2023
39b0cb8
Updated cuda ops for st3d with debugging lines if needed
Jul 14, 2023
c98c8f8
Merge pull request #2 from ut-amrl/st3d++
artzha Jul 14, 2023
a813ea9
exact working steps for conda
sadanand1120 Jul 15, 2023
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
Update README.md
artzha authored Jun 30, 2023
commit a4dd971ac00c0c9c89e6ced51d5ace5435e6c4d0
71 changes: 2 additions & 69 deletions README.md
Original file line number Diff line number Diff line change
@@ -1,72 +1,5 @@
# ST3D & ST3D++

Code release for the paper **ST3D: Self-training for Unsupervised Domain Adaptation on 3D Object Detection**, CVPR 2021 and
**ST3D++: Denoised Self-training for Unsupervised Domain Adaptation on 3D Object Detection**, T-PAMI 2022.


## News
[2022-09-26] ST3D++ (The extension of ST3D) has been integrated in this repo for Waymo->KITTI and nuScenes->KITTI.


## Introduction
Our code is based on OpenPCDet v0.3.0.
More updates on OpenPCDet are supposed to be compatible with our code.

## Model Zoo

### Waymo -> KITTI TASK

| | method | Car@R40 | Ped@R40 | Cyc@R40 |
|---------------------------------------------------------------------------------------------|:--------------:|:-------:|:-------:|:-------:|
| [SECOND-IoU](tools/cfgs/da-waymo-kitti_models/secondiou_st3d/secondiou_st3d_car.yaml) | ST3D | 62.19 | 48.33 | 46.09 |
| [SECOND-IoU](tools/cfgs/da-waymo-kitti_models/secondiou_st3d/secondiou_st3d_car.yaml) | ST3D (w/ sn) | 73.62 | 51.92 | 53.00 |
| [SECOND-IoU](tools/cfgs/da-waymo-kitti_models/secondiou_st3d/secondiou_st3d++_ros_car.yaml) | ST3D++ | 65.10 | 53.87 | 53.43 |
| [SECOND-IoU](tools/cfgs/da-waymo-kitti_models/secondiou_st3d/secondiou_st3d++_sn_car.yaml) | ST3D++ (w/ sn) | 74.73 | 59.21 | 60.76 |
| [PVRCNN](tools/cfgs/da-waymo-kitti_models/pvrcnn_st3d/pvrcnn_st3d.yaml) | ST3D | 64.05 | - | - |
| [PVRCNN](tools/cfgs/da-waymo-kitti_models/pvrcnn_st3d/pvrcnn_st3d.yaml) | ST3D (w/ sn) | 77.33 | - | - |

We could not provide the above pretrained models due to [Waymo Dataset License Agreement](https://waymo.com/open/terms/),
but you should achieve similar performance by training with the default configs. To access these pretrained models, please
send us an email with your name, institute, a screenshot of the Waymo dataset registration confirmation mail, and your
intended usage. Please send a second email if we don't get back to you in two days. Please note that Waymo open dataset is
under strict non-commercial license, so we are not allowed to share the model with you if it will use for any profit-oriented activities.

Also, the training Waymo data used in our work is version 1.0, but the version now available is version 1.2.
The pretrained model on these two version data should be similar when adapted to KITTI.


### nuScenes -> KITTI TASK
| | method | Car@R40 | Ped@R40 | Cyc@R40 |
|------------------------------------------------------------------------------------------------|:--------------:|:-------:|:-------:|:-------:|
| [SECOND-IoU](tools/cfgs/da-nuscenes-kitti_models/secondiou/secondiou_old_anchor.yaml) | Source Only | 17.92 | - | - |
| [SECOND-IoU](tools/cfgs/da-nuscenes-kitti_models/secondiou/secondiou_old_anchor_ros.yaml) | ROS | 25.37 | - | - |
| [SECOND-IoU](tools/cfgs/da-nuscenes-kitti_models/secondiou/secondiou_old_anchor_sn.yaml) | SN | 21.23 | 34.36 | 5.67 |
| [SECOND-IoU](tools/cfgs/da-nuscenes-kitti_models/secondiou_st3d/secondiou_st3d_car.yaml) | ST3D | 55.19 | - | - |
| [SECOND-IoU](tools/cfgs/da-nuscenes-kitti_models/secondiou_st3d/secondiou_st3d_car.yaml) | ST3D (w/ SN) | 62.27 | - | - |
| [SECOND-IoU](tools/cfgs/da-nuscenes-kitti_models/secondiou_st3d/secondiou_st3d++_ros_car.yaml) | ST3D++ | 66.01 | 45.23 | 25.98 |
| [SECOND-IoU](tools/cfgs/da-nuscenes-kitti_models/secondiou_st3d/secondiou_st3d++_sn_car.yaml) | ST3D++ (w/ SN) | 66.24 | 46.75 | 22.66 |
| [PV-RCNN](tools/cfgs/da-nuscenes-kitti_models/pvrcnn/pvrcnn_old_anchor.yaml) | Source Only | 37.17 | - | - |
| [PV-RCNN](tools/cfgs/da-nuscenes-kitti_models/pvrcnn/pvrcnn_old_anchor_ros.yaml) | ROS | 38.84 | - | - |
| [PV-RCNN](tools/cfgs/da-nuscenes-kitti_models/pvrcnn/pvrcnn_old_anchor_sn.yaml) | SN | 49.47 | - | - |
| [PV-RCNN](tools/cfgs/da-nuscenes-kitti_models/pvrcnn_st3d/pvrcnn_st3d.yaml) | ST3D | 71.11 | - | - |
| [PV-RCNN](tools/cfgs/da-nuscenes-kitti_models/pvrcnn_st3d/pvrcnn_st3d.yaml) | ST3D (w/ SN) | 73.16 | - | - |
| [PointRCNN](tools/cfgs/da-nuscenes-kitti_models/pointrcnn/pointrcnn.yaml) | ROS | 55.92 | - | - |
| [PointRCNN](tools/cfgs/da-nuscenes-kitti_models/pointrcnn_st3d/pointrcnn_st3d++_car.yaml) | ST3D++ | 67.51 | - | - |

We provide pretrained models here for nuScenes > KITTI task in [models](https://connecthkuhk-my.sharepoint.com/:f:/g/personal/jhyang13_connect_hku_hk/ErVtcVax3OBJgn4TyQxbOwMBCt1kDCt4_rYaXqHPsg_ZNw?e=a42lgy).


### Waymo -> nuScenes TASK
| | method | Car@R11 | Car@R40 | download |
|---------------------------------------------|----------:|:-------:|:-------:|:---------:|
| [SECOND-IoU](tools/cfgs/da-waymo-nus_models/secondiou_st3d/secondiou_st3d.yaml) | ST3D | 23.24 | 20.19 | [model](https://drive.google.com/file/d/150gvNE29iZ8Ah4L4uO_E40x0fyr_cn_J/view?usp=sharing) |
| [PVRCNN](tools/cfgs/da-waymo-nus_models/pvrcnn_st3d/pvrcnn_st3d.yaml) | ST3D | 27.18 | 22.99 | [model](https://drive.google.com/file/d/1Gk5i4qRz5tLU7Yd4U-oPureN9cV6TaLa/view?usp=sharing) |
| [SECOND-IoU](tools/cfgs/da-waymo-nus_models/secondiou_st3d/secondiou_st3d.yaml) | ST3D (w/ sn) | 23.52 | 20.38 | [model](https://drive.google.com/file/d/1WUHBTZylVTFLdSvO8wrgti27SEjFJxQl/view?usp=sharing) |
| [PVRCNN](tools/cfgs/da-waymo-nus_models/pvrcnn_st3d/pvrcnn_st3d.yaml) | ST3D (w/ sn) | 28.06 | 23.67 | [model](https://drive.google.com/file/d/1KTCvmltfwIEh9i2t3Pu9APp_2GKyENme/view?usp=sharing) |

We could not provide the above pretrained models due to [Waymo Dataset License Agreement](https://waymo.com/open/terms/),
but you should achieve similar performance by training with the default configs.

# UT CODa Model Training Configurations
We utilize the ST3D++ code repository and OpenPCDet model training configuration files for all empirical analyses conducted for the UT CODa paper.

## Installation