Skip to content
Snippets Groups Projects
Commit e79aced7 authored by zhangwenwei's avatar zhangwenwei
Browse files

Merge branch 'clean-ci' into 'master' 

modify readme and changelog

See merge request open-mmlab/mmdet.3d!145
parents 0e60e20d f0b23ecb
Branches
No related tags found
No related merge requests found
Show whitespace changes
Inline Side-by-side
.travis.yml deleted 100644 → 0
+ 0
− 43
View file @ 0e60e20d
dist: bionic # ubuntu 18.04
language: python
python:
- "3.5"
- "3.6"
- "3.7"
env: CUDA=10.1.105-1 CUDA_SHORT=10.1 UBUNTU_VERSION=ubuntu1804 FORCE_CUDA=1
cache: pip
# Ref to CUDA installation in Travis: https://github.com/jeremad/cuda-travis
before_install:
- INSTALLER=cuda-repo-${UBUNTU_VERSION}_${CUDA}_amd64.deb
- wget http://developer.download.nvidia.com/compute/cuda/repos/${UBUNTU_VERSION}/x86_64/${INSTALLER}
- sudo dpkg -i ${INSTALLER}
- wget https://developer.download.nvidia.com/compute/cuda/repos/${UBUNTU_VERSION}/x86_64/7fa2af80.pub
- sudo apt-key add 7fa2af80.pub
- sudo apt update -qq
- sudo apt install -y cuda-${CUDA_SHORT/./-} cuda-cufft-dev-${CUDA_SHORT/./-}
- sudo apt clean
- CUDA_HOME=/usr/local/cuda-${CUDA_SHORT}
- LD_LIBRARY_PATH=${CUDA_HOME}/lib64:${CUDA_HOME}/include:${LD_LIBRARY_PATH}
- PATH=${CUDA_HOME}/bin:${PATH}
install:
- pip install Pillow==6.2.2 # remove this line when torchvision>=0.5
- pip install torch==1.2 torchvision==0.4.0 # TODO: fix CI for pytorch>1.2
- pip install "git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI"
- pip install -r requirements.txt
before_script:
- flake8 .
- isort -rc --check-only --diff mmdet3d/ tools/ tests/
- yapf -r -d --style .style.yapf mmdet3d/ tools/ tests/ configs/
script:
- python setup.py check -m -s
- python setup.py build_ext --inplace
- coverage run --source mmdet3d -m py.test -v --xdoctest-modules tests mmdet3d
after_success:
- coverage report
README.md
+ 9
− 11
View file @ e79aced7
...@@ -10,7 +10,7 @@ Documentation: https://mmdetection3d.readthedocs.io/ ...@@ -10,7 +10,7 @@ Documentation: https://mmdetection3d.readthedocs.io/
The master branch works with **PyTorch 1.3 to 1.5**. The master branch works with **PyTorch 1.3 to 1.5**.
MMDetection3D is an open source object detection toolbox based on PyTorch, towards the next generation toolbox for general 3D detection. It is MMDetection3D is an open source object detection toolbox based on PyTorch, towards the next-generation platform for general 3D detection. It is
a part of the OpenMMLab project developed by [MMLab](http://mmlab.ie.cuhk.edu.hk/). a part of the OpenMMLab project developed by [MMLab](http://mmlab.ie.cuhk.edu.hk/).
![demo image](resources/outdoor_demo.gif) ![demo image](resources/outdoor_demo.gif)
...@@ -19,20 +19,21 @@ a part of the OpenMMLab project developed by [MMLab](http://mmlab.ie.cuhk.edu.hk ...@@ -19,20 +19,21 @@ a part of the OpenMMLab project developed by [MMLab](http://mmlab.ie.cuhk.edu.hk
- **Support multi-modality/single-modality detectors out of box** - **Support multi-modality/single-modality detectors out of box**
The toolbox directly supports multi-modality/single-modality detectors including MVXNet, VoteNet, PointPillars, etc. It directly supports multi-modality/single-modality detectors including MVXNet, VoteNet, PointPillars, etc.
- **Support indoor/outdoor 3D detection out of box** - **Support indoor/outdoor 3D detection out of box**
The toolbox directly supports popular indoor and outdoor 3D detection datasets, including ScanNet, SUNRGB-D, nuScenes, Lyft, and KITTI. It directly supports popular indoor and outdoor 3D detection datasets, including ScanNet, SUNRGB-D, nuScenes, Lyft, and KITTI.
- **Natural integration with 2D detection** - **Natural integration with 2D detection**
All the about **300 models, 40+ papers**, and modules supported in [MMDetection's model zoo](https://github.com/open-mmlab/mmdetection/blob/master/docs/model_zoo.md) can be trained or used in this codebase.
All the about **300 models, methods of 40+ papers**, and modules supported in [MMDetection](https://github.com/open-mmlab/mmdetection/blob/master/docs/model_zoo.md) can be trained or used in this codebase.
- **High efficiency** - **High efficiency**
The training and testing speed is [faster than other codebases](./docs/benchmarks.md). It trains [faster than other codebases](./docs/benchmarks.md).
Apart from MMDetection3D, we also released a library [MMDetection](https://github.com/open-mmlab/mmdetection) and [mmcv](https://github.com/open-mmlab/mmcv) for computer vision research, which are heavily depended on by this toolbox. Apart from MMDetection3D, we also released a library [MMDetection](https://github.com/open-mmlab/mmdetection) and [MMCV](https://github.com/open-mmlab/mmcv) for computer vision research, which are heavily depended on by this toolbox. Like [MMDetection](https://github.com/open-mmlab/mmdetection) and [MMCV](https://github.com/open-mmlab/mmcv), MMDetection3D can also be used as a library to support different projects on top of it.
## License ## License
...@@ -40,7 +41,7 @@ This project is released under the [Apache 2.0 license](LICENSE). ...@@ -40,7 +41,7 @@ This project is released under the [Apache 2.0 license](LICENSE).
## Changelog ## Changelog
v0.1.0 was released in 8/7/2020. v0.1.0 was released in 9/7/2020.
Please refer to [changelog.md](docs/changelog.md) for details and release history. Please refer to [changelog.md](docs/changelog.md) for details and release history.
## Benchmark and model zoo ## Benchmark and model zoo
...@@ -60,13 +61,12 @@ Results and models are available in the [model zoo](docs/model_zoo.md). ...@@ -60,13 +61,12 @@ Results and models are available in the [model zoo](docs/model_zoo.md).
Other features Other features
- [x] [Dynamic Voxelization](configs/carafe/README.md) - [x] [Dynamic Voxelization](configs/carafe/README.md)
**Note:** All the about **300 models, 40+ papers**, and modules supported in [MMDetection's model zoo](https://github.com/open-mmlab/mmdetection/blob/master/docs/model_zoo.md) can be trained or used in this codebase. **Note:** All the about **300 models, methods of 40+ papers** in 2D detection supported by [MMDetection](https://github.com/open-mmlab/mmdetection/blob/master/docs/model_zoo.md) can be trained or used in this codebase.
## Installation ## Installation
Please refer to [install.md](docs/install.md) for installation and dataset preparation. Please refer to [install.md](docs/install.md) for installation and dataset preparation.
## Get Started ## Get Started
Please see [getting_started.md](docs/getting_started.md) for the basic usage of MMDetection. There are also tutorials for [finetuning models](docs/tutorials/finetune.md), [adding new dataset](docs/tutorials/new_dataset.md), [designing data pipeline](docs/tutorials/data_pipeline.md), and [adding new modules](docs/tutorials/new_modules.md). Please see [getting_started.md](docs/getting_started.md) for the basic usage of MMDetection. There are also tutorials for [finetuning models](docs/tutorials/finetune.md), [adding new dataset](docs/tutorials/new_dataset.md), [designing data pipeline](docs/tutorials/data_pipeline.md), and [adding new modules](docs/tutorials/new_modules.md).
...@@ -80,7 +80,6 @@ We appreciate all contributions to improve MMDetection3D. Please refer to [CONTR ...@@ -80,7 +80,6 @@ We appreciate all contributions to improve MMDetection3D. Please refer to [CONTR
MMDetection3D is an open source project that is contributed by researchers and engineers from various colleges and companies. We appreciate all the contributors as well as users who give valuable feedbacks. MMDetection3D is an open source project that is contributed by researchers and engineers from various colleges and companies. We appreciate all the contributors as well as users who give valuable feedbacks.
We wish that the toolbox and benchmark could serve the growing research community by providing a flexible toolkit to reimplement existing methods and develop their own new 3D detectors. We wish that the toolbox and benchmark could serve the growing research community by providing a flexible toolkit to reimplement existing methods and develop their own new 3D detectors.
## Citation ## Citation
If you use this toolbox or benchmark in your research, please cite this project. If you use this toolbox or benchmark in your research, please cite this project.
...@@ -96,7 +95,6 @@ If you use this toolbox or benchmark in your research, please cite this project. ...@@ -96,7 +95,6 @@ If you use this toolbox or benchmark in your research, please cite this project.
} }
``` ```
## Contact ## Contact
This repo is currently maintained by Wenwei Zhang ([@ZwwWayne](https://github.com/ZwwWayne)), Yuefeng Wu ([@xavierwu95](https://github.com/xavierwu95)), Tai Wang ([@Tai-Wang](https://github.com/Tai-Wang)), and Yinhao Li ([@yinchimaoliang](https://github.com/yinchimaoliang)). This repo is currently maintained by Wenwei Zhang ([@ZwwWayne](https://github.com/ZwwWayne)), Yuefeng Wu ([@xavierwu95](https://github.com/xavierwu95)), Tai Wang ([@Tai-Wang](https://github.com/Tai-Wang)), and Yinhao Li ([@yinchimaoliang](https://github.com/yinchimaoliang)).
docs/benchmarks.md
+ 12
− 248
View file @ e79aced7
...@@ -10,86 +10,18 @@ with some other open source 3D detection codebases. ...@@ -10,86 +10,18 @@ with some other open source 3D detection codebases.
* Software: Python 3.7, CUDA 10.1, cuDNN 7.6.5, PyTorch 1.3, numba 0.48.0. * Software: Python 3.7, CUDA 10.1, cuDNN 7.6.5, PyTorch 1.3, numba 0.48.0.
* Model: Since all the other codebases implements different models, we compare the corresponding models including SECOND, PointPillars, Part-A2, and VoteNet with them separately. * Model: Since all the other codebases implements different models, we compare the corresponding models including SECOND, PointPillars, Part-A2, and VoteNet with them separately.
* Metrics: We use the average throughput in iterations of the entire training run and skip the first 50 iterations of each epoch to skip GPU warmup time. * Metrics: We use the average throughput in iterations of the entire training run and skip the first 50 iterations of each epoch to skip GPU warmup time.
For `single GPU inference speed`, we calculate the FPS in 2000 iterations after 5 warmup iterations.
## Main Results ## Main Results
### VoteNet We compare the training speed (samples/s) with other codebases if they implement the similar models. The results are as below, the greater the numbers in the table, the faster of the training process.
We compare our implementation of VoteNet with [votenet](https://github.com/facebookresearch/votenet/) and report the performance on SUNRGB-D v2 dataset under the AP@0.5 metric. We find that our implementation achieves higher accuracy, so we also report the AP here.
```eval_rst
+----------------+---------------------+--------------------+--------+
| Implementation | Training (sample/s) | Testing (sample/s) | AP@0.5 |
+================+=====================+====================+========+
| MMDetection3D | 358 | 17 | 35.8 |
+----------------+---------------------+--------------------+--------+
| votenet | 77 | 3 | 31.5 |
+----------------+---------------------+--------------------+--------+
```
### Single-Class PointPillars
Since [Det3D](https://github.com/poodarchu/Det3D/) only provides PointPillars on car class, we compare the training speed of single-class PointPillars here.
```eval_rst
+----------------+---------------------+--------------------+
| Implementation | Training (sample/s) | Testing (sample/s) |
+================+=====================+====================+
| MMDetection3D | 141 | 44 |
+----------------+---------------------+--------------------+
| Det3D | 140 | 24 |
+----------------+---------------------+--------------------+
```
### Multi-Class PointPillars
[OpenPCDet](https://github.com/open-mmlab/OpenPCDet/tree/b32fbddbe06183507bad433ed99b407cbc2175c2) only provides PointPillars
on 3 classes, we compare the training speed of multi-class PointPillars here.
Note that we reimplement voxelization process on GPU using PyTorch, so the voxelization time is taken into count, however, other codebases apply voxelization in the data preprocessing and do not take this time into FPS calculation. Therefore we report two inference speed, with or without the voxelization time and compare with other codebases without calculating voxelization in the column of ``Calibrated Testing``.
```eval_rst | Methods | MMDetection3D |votenet| Det3D | OpenPCDet |
+----------------+---------------------+--------------------+-------------------------------+ |:-------:|:-------------:|:-----:|:-----:|:---------:|
| Implementation | Training (sample/s) | Testing (sample/s) | Calibrated Testing (sample/s) | | VoteNet | 358 | 77 | × | × |
+================+=====================+====================+===============================+ | PointPillars-car| 141 | × | 140 | × |
| MMDetection3D | 107 | 45 | 65 | | PointPillars-3class| 107 | × | × |44 |
+----------------+---------------------+--------------------+-------------------------------+ | SECOND | 40 | × | × |30 |
| OpenPCDet | 44 | - | 59 | | Part-A2| 17 | × | × |14 |
+----------------+---------------------+--------------------+-------------------------------+
```
### SECOND
[Det3D](https://github.com/poodarchu/Det3D/) provides a different SECOND on car class and we cannot train the original SECOND by modifying the config.
So we only compare SECOND with [OpenPCDet](https://github.com/open-mmlab/OpenPCDet/tree/b32fbddbe06183507bad433ed99b407cbc2175c2), which is a SECOND model on 3 classes, we report the AP on moderate
condition following the KITTI benchmark and compare average AP over all classes on moderate condition for
performance on 3 classes.
```eval_rst
+----------------+---------------------+--------------------+-------------------------------+
| Implementation | Training (sample/s) | Testing (sample/s) | Calibrated Testing (sample/s) |
+================+=====================+====================+===============================+
| MMDetection3D | 40 | 25 | 30 |
+----------------+---------------------+--------------------+-------------------------------+
| OpenPCDet | 30 | - | 27 |
+----------------+---------------------+--------------------+-------------------------------+
```
### Part-A2
We benchmark Part-A2 with that in [OpenPCDet](https://github.com/open-mmlab/OpenPCDet/tree/b32fbddbe06183507bad433ed99b407cbc2175c2). We report the AP on moderate condition following the KITTI benchmark
and compare average AP over all classes on moderate condition for performance on 3 classes.
```eval_rst
+----------------+---------------------+--------------------+-------------------------------+
| Implementation | Training (sample/s) | Testing (sample/s) | Calibrated Testing (sample/s) |
+================+=====================+====================+===============================+
| MMDetection3D | 17 | 16 | 18 |
+----------------+---------------------+--------------------+-------------------------------+
| OpenPCDet | 14 | - | 12 |
+----------------+---------------------+--------------------+-------------------------------+
```
## Details of Comparison ## Details of Comparison
...@@ -183,58 +115,6 @@ and compare average AP over all classes on moderate condition for performance on ...@@ -183,58 +115,6 @@ and compare average AP over all classes on moderate condition for performance on
</details> </details>
For testing speed, we add code to record the running time in the file `./tools/eval_utils/eval_utils.py`.
<details>
<summary>
(diff to make it use the same method for benchmarking speed - click to expand)
</summary>
```diff
diff --git a/tools/eval_utils/eval_utils.py b/tools/eval_utils/eval_utils.py
index 0cbf17b..f51e687 100644
--- a/tools/eval_utils/eval_utils.py
+++ b/tools/eval_utils/eval_utils.py
@@ -49,8 +49,11 @@ def eval_one_epoch(cfg, model, dataloader, epoch_id, logger, dist_test=False, sa
if cfg.LOCAL_RANK == 0:
progress_bar = tqdm.tqdm(total=len(dataloader), leave=True, desc='eval', dynamic_ncols=True)
- start_time = time.time()
+ num_warmup = 5
+ pure_inf_time = 0
for i, batch_dict in enumerate(dataloader):
+ torch.cuda.synchronize()
+ start_time = time.perf_counter()
for key, val in batch_dict.items():
if not isinstance(val, np.ndarray):
continue
@@ -61,7 +64,14 @@ def eval_one_epoch(cfg, model, dataloader, epoch_id, logger, dist_test=False, sa
with torch.no_grad():
pred_dicts, ret_dict = model(batch_dict)
disp_dict = {}
-
+ torch.cuda.synchronize()
+ elapsed = time.perf_counter() - start_time
+ if i >= num_warmup:
+ pure_inf_time += elapsed
+ if (i + 1) == 2000:
+ pure_inf_time += elapsed
+ fps = (i + 1 - num_warmup) / pure_inf_time
+ out_str = f'Overall fps: {fps:.1f} img / s'
statistics_info(cfg, ret_dict, metric, disp_dict)
annos = dataset.generate_prediction_dicts(
batch_dict, pred_dicts, class_names,
@@ -71,7 +81,7 @@ def eval_one_epoch(cfg, model, dataloader, epoch_id, logger, dist_test=False, sa
if cfg.LOCAL_RANK == 0:
progress_bar.set_postfix(disp_dict)
progress_bar.update()
-
+ print(out_str)
if cfg.LOCAL_RANK == 0:
progress_bar.close()
```
</details>
### VoteNet ### VoteNet
* __MMDetection3D__: With release v0.1.0, run * __MMDetection3D__: With release v0.1.0, run
...@@ -243,25 +123,13 @@ and compare average AP over all classes on moderate condition for performance on ...@@ -243,25 +123,13 @@ and compare average AP over all classes on moderate condition for performance on
./tools/dist_train.sh configs/votenet/votenet_16x8_sunrgbd-3d-10class.py 8 --no-validate ./tools/dist_train.sh configs/votenet/votenet_16x8_sunrgbd-3d-10class.py 8 --no-validate
``` ```
Then benchmark the test speed by running * __votenet__: At commit [2f6d6d3](https://github.com/facebookresearch/votenet/tree/2f6d6d36ff98d96901182e935afe48ccee82d566), run
```bash
python tools/benchmark.py configs/votenet/votenet_16x8_sunrgbd-3d-10class.py ${CHECKPOINTS}
```
* __votenet__: At commit 2f6d6d3, run
```bash ```bash
python train.py --dataset sunrgbd --batch_size 16 python train.py --dataset sunrgbd --batch_size 16
``` ```
Then benchmark the test speed by running ### PointPillars-car
```bash
python eval.py --dataset sunrgbd --checkpoint_path log_sunrgbd/checkpoint.tar --batch_size 1 --dump_dir eval_sunrgbd --cluster_sampling seed_fps --use_3d_nms --use_cls_nms --per_class_proposal
```
### Single-class PointPillars
* __MMDetection3D__: With release v0.1.0, run * __MMDetection3D__: With release v0.1.0, run
...@@ -269,13 +137,7 @@ and compare average AP over all classes on moderate condition for performance on ...@@ -269,13 +137,7 @@ and compare average AP over all classes on moderate condition for performance on
./tools/dist_train.sh configs/benchmark/hv_pointpillars_secfpn_3x8_100e_det3d_kitti-3d-car.py 8 --no-validate ./tools/dist_train.sh configs/benchmark/hv_pointpillars_secfpn_3x8_100e_det3d_kitti-3d-car.py 8 --no-validate
``` ```
Then benchmark the test speed by running * __Det3D__: At commit [519251e](https://github.com/poodarchu/Det3D/tree/519251e72a5c1fdd58972eabeac67808676b9bb7), use `kitti_point_pillars_mghead_syncbn.py` and run
```bash
python tools/benchmark.py configs/benchmark/hv_pointpillars_secfpn_3x8_100e_det3d_kitti-3d-car.py ${CHECKPOINT}
```
* __Det3D__: At commit 519251e, use kitti_point_pillars_mghead_syncbn.py and run
```bash ```bash
./tools/scripts/train.sh --launcher=slurm --gpus=8 ./tools/scripts/train.sh --launcher=slurm --gpus=8
...@@ -309,66 +171,7 @@ and compare average AP over all classes on moderate condition for performance on ...@@ -309,66 +171,7 @@ and compare average AP over all classes on moderate condition for performance on
</details> </details>
Then benchmark the test speed by running ### PointPillars-3class
```bash
./tools/scripts/test.sh examples/point_pillars/configs/kitti_point_pillars_mghead_syncbn.py ./work_dir/Point_Pillars/latest.pth
```
Note that the `tools/dist_test.py` is modified to benchmark point pillars.
<details>
<summary>
(diff to benchmark the similar models - click to expand)
</summary>
```diff
diff --git a/tools/dist_test.py b/tools/dist_test.py
index 3e37f8a..0908fee 100644
--- a/tools/dist_test.py
+++ b/tools/dist_test.py
@@ -3,6 +3,7 @@ import json
import os
import sys
+import time
import apex
import numpy as np
import torch
@@ -128,12 +129,26 @@ def main():
detections = {}
cpu_device = torch.device("cpu")
+ sample_time = list()
for i, data_batch in enumerate(data_loader):
with torch.no_grad():
+
+ torch.cuda.synchronize()
+ start_time = time.perf_counter()
+
outputs = batch_processor(
model, data_batch, train_mode=False, local_rank=args.local_rank,
)
+
+ torch.cuda.synchronize()
+ elapsed = time.perf_counter() - start_time
+ sample_time.append(elapsed)
+ if i == 2006:
+ st_arr = np.array(sample_time)[5:]
+
+ print('avg time elapsed: %f s'%st_arr.mean())
+
for output in outputs:
token = output["metadata"]["token"]
for k, v in output.items():
@@ -185,3 +200,4 @@ def main():
if __name__ == "__main__":
main()
```
### Multi-class PointPillars
* __MMDetection3D__: With release v0.1.0, run * __MMDetection3D__: With release v0.1.0, run
...@@ -376,12 +179,6 @@ and compare average AP over all classes on moderate condition for performance on ...@@ -376,12 +179,6 @@ and compare average AP over all classes on moderate condition for performance on
./tools/dist_train.sh configs/benchmark/hv_pointpillars_secfpn_4x8_80e_pcdet_kitti-3d-3class.py 8 --no-validate ./tools/dist_train.sh configs/benchmark/hv_pointpillars_secfpn_4x8_80e_pcdet_kitti-3d-3class.py 8 --no-validate
``` ```
Then benchmark the test speed by running
```bash
python tools/benchmark.py configs/benchmark/hv_pointpillars_secfpn_4x8_80e_pcdet_kitti-3d-3class.py ${CKPT}
```
* __OpenPCDet__: At commit [b32fbddb](https://github.com/open-mmlab/OpenPCDet/tree/b32fbddbe06183507bad433ed99b407cbc2175c2), run * __OpenPCDet__: At commit [b32fbddb](https://github.com/open-mmlab/OpenPCDet/tree/b32fbddbe06183507bad433ed99b407cbc2175c2), run
```bash ```bash
...@@ -389,13 +186,6 @@ and compare average AP over all classes on moderate condition for performance on ...@@ -389,13 +186,6 @@ and compare average AP over all classes on moderate condition for performance on
sh scripts/slurm_train.sh ${PARTITION} ${JOB_NAME} 8 --cfg_file ./cfgs/pointpillar.yaml --batch_size 32 --workers 32 sh scripts/slurm_train.sh ${PARTITION} ${JOB_NAME} 8 --cfg_file ./cfgs/pointpillar.yaml --batch_size 32 --workers 32
``` ```
Then benchmark the test speed by running
```bash
cd tools
python test.py --cfg_file cfgs/kitti_models/pointpillar.yaml --ckpt ${CKPT} --batch_size 1 --workers 4
```
### SECOND ### SECOND
* __MMDetection3D__: With release v0.1.0, run * __MMDetection3D__: With release v0.1.0, run
...@@ -404,12 +194,6 @@ and compare average AP over all classes on moderate condition for performance on ...@@ -404,12 +194,6 @@ and compare average AP over all classes on moderate condition for performance on
./tools/dist_train.sh configs/benchmark/hv_second_secfpn_4x8_80e_pcdet_kitti-3d-3class.py 8 --no-validate ./tools/dist_train.sh configs/benchmark/hv_second_secfpn_4x8_80e_pcdet_kitti-3d-3class.py 8 --no-validate
``` ```
Then benchmark the test speed by running
```bash
python tools/benchmark.py configs/benchmark/hv_second_secfpn_4x8_80e_pcdet_kitti-3d-3class.py ${CKPT}
```
* __OpenPCDet__: At commit [b32fbddb](https://github.com/open-mmlab/OpenPCDet/tree/b32fbddbe06183507bad433ed99b407cbc2175c2), run * __OpenPCDet__: At commit [b32fbddb](https://github.com/open-mmlab/OpenPCDet/tree/b32fbddbe06183507bad433ed99b407cbc2175c2), run
```bash ```bash
...@@ -417,13 +201,6 @@ and compare average AP over all classes on moderate condition for performance on ...@@ -417,13 +201,6 @@ and compare average AP over all classes on moderate condition for performance on
./scripts/slurm_train.sh ${PARTITION} ${JOB_NAME} 8 --cfg_file ./cfgs/second.yaml --batch_size 32 --workers 32 ./scripts/slurm_train.sh ${PARTITION} ${JOB_NAME} 8 --cfg_file ./cfgs/second.yaml --batch_size 32 --workers 32
``` ```
Then benchmark the test speed by running
```bash
cd tools
python test.py --cfg_file cfgs/kitti_models/second.yaml --ckpt ${CKPT} --batch_size 1 --workers 4
```
### Part-A2 ### Part-A2
* __MMDetection3D__: With release v0.1.0, run * __MMDetection3D__: With release v0.1.0, run
...@@ -432,22 +209,9 @@ and compare average AP over all classes on moderate condition for performance on ...@@ -432,22 +209,9 @@ and compare average AP over all classes on moderate condition for performance on
./tools/dist_train.sh configs/benchmark/hv_PartA2_secfpn_4x8_cyclic_80e_pcdet_kitti-3d-3class.py 8 --no-validate ./tools/dist_train.sh configs/benchmark/hv_PartA2_secfpn_4x8_cyclic_80e_pcdet_kitti-3d-3class.py 8 --no-validate
``` ```
Then benchmark the test speed by running
```bash
python tools/benchmark.py configs/benchmark/hv_PartA2_secfpn_4x8_cyclic_80e_pcdet_kitti-3d-3class.py ${CKPT}
```
* __OpenPCDet__: At commit [b32fbddb](https://github.com/open-mmlab/OpenPCDet/tree/b32fbddbe06183507bad433ed99b407cbc2175c2), train the model by running * __OpenPCDet__: At commit [b32fbddb](https://github.com/open-mmlab/OpenPCDet/tree/b32fbddbe06183507bad433ed99b407cbc2175c2), train the model by running
```bash ```bash
cd tools cd tools
./scripts/slurm_train.sh ${PARTITION} ${JOB_NAME} 8 --cfg_file ./cfgs/PartA2.yaml --batch_size 32 --workers 32 ./scripts/slurm_train.sh ${PARTITION} ${JOB_NAME} 8 --cfg_file ./cfgs/PartA2.yaml --batch_size 32 --workers 32
``` ```
Then benchmark the test speed by running
```bash
cd tools
python test.py --cfg_file cfgs/kitti_models/PartA2.yaml --ckpt ${CKPT} --batch_size 1 --workers 4
```
docs/changelog.md
+ 1
− 1
View file @ e79aced7
## Changelog ## Changelog
### v0.1.0 (8/7/2020) ### v0.1.0 (9/7/2020)
MMDetection3D is released. MMDetection3D is released.
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Please register or to comment