monocon 环境配置详细步骤

提示：文章写完后，目录可以自动生成，如何生成可参考右边的帮助文档

---

前言

3D 目标检测是

---

一、monocon简介

monocon 是一个延续CenterNet框架的3D 目标检测网络；在不依赖dcn 模块的情况下有不错的性能。

二、环境配置

1.下载

代码如下（示例）：

git clone https://github.com/2gunsu/monocon-pytorch.git

cd monocon-pytorch

2.创建环境

conda create -n monocon-pytorch python=3.8
conda activate monocon-pytorch

3、安装其他包

pip install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 --index-url https://download.pytorch.org/whl/cu118

pip install -r requirements.txt 

三、遇到问题

1、AttributeError: module ‘distutils’ has no attribute ‘version’

python train.py
Traceback (most recent call last):
  File "train.py", line 6, in <module>
    from engine.monocon_engine import MonoconEngine
  File "/devdata/deeplearn/cv/3D/monocon-pytorch/engine/monocon_engine.py", line 14, in <module>
    from engine.base_engine import BaseEngine
  File "/devdata/deeplearn/cv/3D/monocon-pytorch/engine/base_engine.py", line 11, in <module>
    from torch.utils.tensorboard import SummaryWriter
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/torch/utils/tensorboard/__init__.py", line 4, in <module>
    LooseVersion = distutils.version.LooseVersion
AttributeError: module 'distutils' has no attribute 'version'

解决：

方法一：将setuptools版本降低到59.5.0
# If you use pip:
pip install setuptools==59.5.0

# For pip3:
pip3 install setuptools==59.5.0

# If you use conda:
conda install setuptools=59.5.0

方法二：升级或者安装更高版本的torch
# If you use pip:
pip install torch==1.11.0

# For pip3:
pip3 install torch==1.11.0

# If you use conda:
conda install pytorch=1.11.0

2、libNVVM cannot be found. Do conda install cudatoolkit:

python train.py
Traceback (most recent call last):
  File "train.py", line 6, in <module>
    from engine.monocon_engine import MonoconEngine
  File "/devdata/deeplearn/cv/3D/monocon-pytorch/engine/monocon_engine.py", line 15, in <module>
    from dataset.monocon_dataset import MonoConDataset
  File "/devdata/deeplearn/cv/3D/monocon-pytorch/dataset/monocon_dataset.py", line 11, in <module>
    from dataset.base_dataset import BaseKITTIMono3DDataset
  File "/devdata/deeplearn/cv/3D/monocon-pytorch/dataset/base_dataset.py", line 12, in <module>
    from engine.kitti_eval import kitti_eval
  File "/devdata/deeplearn/cv/3D/monocon-pytorch/engine/kitti_eval/__init__.py", line 1, in <module>
    from .eval import kitti_eval, do_eval
  File "/devdata/deeplearn/cv/3D/monocon-pytorch/engine/kitti_eval/eval.py", line 11, in <module>
    from kitti_eval.rotate_iou import rotate_iou_gpu_eval
  File "/devdata/deeplearn/cv/3D/monocon-pytorch/engine/kitti_eval/../kitti_eval/__init__.py", line 1, in <module>
    from .eval import kitti_eval, do_eval
  File "/devdata/deeplearn/cv/3D/monocon-pytorch/engine/kitti_eval/../kitti_eval/eval.py", line 11, in <module>
    from kitti_eval.rotate_iou import rotate_iou_gpu_eval
  File "/devdata/deeplearn/cv/3D/monocon-pytorch/engine/kitti_eval/../kitti_eval/rotate_iou.py", line 283, in <module>
    def rotate_iou_kernel_eval(N,
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/cuda/decorators.py", line 115, in _jit
    disp.compile(argtypes)
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/cuda/dispatcher.py", line 794, in compile
    kernel = _Kernel(self.py_func, argtypes, **self.targetoptions)
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/core/compiler_lock.py", line 35, in _acquire_compile_lock
    return func(*args, **kwargs)
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/cuda/dispatcher.py", line 75, in __init__
    cres = compile_cuda(self.py_func, types.void, self.argtypes,
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/core/compiler_lock.py", line 35, in _acquire_compile_lock
    return func(*args, **kwargs)
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/cuda/compiler.py", line 210, in compile_cuda
    cres = compiler.compile_extra(typingctx=typingctx,
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/core/compiler.py", line 716, in compile_extra
    return pipeline.compile_extra(func)
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/core/compiler.py", line 452, in compile_extra
    return self._compile_bytecode()
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/core/compiler.py", line 520, in _compile_bytecode
    return self._compile_core()
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/core/compiler.py", line 499, in _compile_core
    raise e
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/core/compiler.py", line 486, in _compile_core
    pm.run(self.state)
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/core/compiler_machinery.py", line 368, in run
    raise patched_exception
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/core/compiler_machinery.py", line 356, in run
    self._runPass(idx, pass_inst, state)
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/core/compiler_lock.py", line 35, in _acquire_compile_lock
    return func(*args, **kwargs)
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/core/compiler_machinery.py", line 311, in _runPass
    mutated |= check(pss.run_pass, internal_state)
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/core/compiler_machinery.py", line 273, in check
    mangled = func(compiler_state)
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/core/typed_passes.py", line 105, in run_pass
    typemap, return_type, calltypes, errs = type_inference_stage(
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/core/typed_passes.py", line 83, in type_inference_stage
    errs = infer.propagate(raise_errors=raise_errors)
  File "/devdata/anaconda3/envs/monocon-pytorch/lib/python3.8/site-packages/numba/core/typeinfer.py", line 1086, in propagate
    raise errors[0]
numba.core.errors.TypingError: Failed in cuda mode pipeline (step: nopython frontend)
Failed in cuda mode pipeline (step: nopython frontend)
Failed in cuda mode pipeline (step: nopython frontend)
Internal error at <numba.core.typeinfer.CallConstraint object at 0x7f2e00c0a8b0>.
libNVVM cannot be found. Do `conda install cudatoolkit`:
[Errno 2] No such file or directory: '/usr/local/cuda-11.8:/nvvm/lib64'
During: resolving callee type: type(CUDADispatcher(<function rbbox_to_corners at 0x7f2e01e80550>))
During: typing of call at /devdata/deeplearn/cv/3D/monocon-pytorch/engine/kitti_eval/../kitti_eval/rotate_iou.py (241)

Enable logging at debug level for details.

File "engine/kitti_eval/rotate_iou.py", line 241:
def inter(rbbox1, rbbox2):
    <source elided>

    rbbox_to_corners(corners1, rbbox1)
    ^

During: resolving callee type: type(CUDADispatcher(<function inter at 0x7f2e01e51700>))
During: typing of call at /devdata/deeplearn/cv/3D/monocon-pytorch/engine/kitti_eval/../kitti_eval/rotate_iou.py (269)


File "engine/kitti_eval/rotate_iou.py", line 269:
def devRotateIoUEval(rbox1, rbox2, criterion=-1):
    <source elided>
    area2 = rbox2[2] * rbox2[3]
    area_inter = inter(rbox1, rbox2)
    ^

During: resolving callee type: type(CUDADispatcher(<function devRotateIoUEval at 0x7f2e01e518b0>))
During: typing of call at /devdata/deeplearn/cv/3D/monocon-pytorch/engine/kitti_eval/../kitti_eval/rotate_iou.py (332)


File "engine/kitti_eval/rotate_iou.py", line 332:
def rotate_iou_kernel_eval(N,
    <source elided>
                tx * K + i)
            dev_iou[offset] = devRotateIoUEval(block_qboxes[i * 5:i * 5 + 5],

解决办法：

pip install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 --index-url https://download.pytorch.org/whl/cu118
conda install cudatoolkit==11.8 -c nvidia //安装对应版本的cuda

3、解压kitti 3D 的数据遇到error: invalid zip file with overlapped components (possible zip bomb)

解决办法：安装7z 来解压

sudo apt-get install p7zip
sudo apt-get install p7zip-full
sudo apt-get install p7zip-rar

7z x 001.zip //001.zip 是需要解压的文件

---

四、训练结果

1、训练

python train.py

2、结果

[2024-05-15 06:03:10] Evaluating on Epoch 200...
Collecting Results...: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 472/472 [04:16<00:00,  1.84it/s]

----------- Eval Results ------------
Pedestrian AP40@0.50, 0.50, 0.50:
bbox AP40:67.7988, 54.6922, 47.5099
bev  AP40:2.6367, 2.3894, 1.8246
3d   AP40:1.9083, 1.6483, 1.5825
aos  AP40:55.60, 44.47, 38.21
Pedestrian AP40@0.50, 0.25, 0.25:
bbox AP40:67.7988, 54.6922, 47.5099
bev  AP40:16.7131, 14.1143, 11.7589
3d   AP40:15.5213, 13.6702, 11.4670
aos  AP40:55.60, 44.47, 38.21
Cyclist AP40@0.50, 0.50, 0.50:
bbox AP40:63.4784, 37.8742, 34.3940
bev  AP40:6.7544, 3.7150, 3.0675
3d   AP40:5.1372, 2.6349, 2.6652
aos  AP40:58.13, 34.36, 31.23
Cyclist AP40@0.50, 0.25, 0.25:
bbox AP40:63.4784, 37.8742, 34.3940
bev  AP40:24.8584, 14.0738, 12.7809
3d   AP40:22.3357, 12.2825, 11.7030
aos  AP40:58.13, 34.36, 31.23
Car AP40@0.70, 0.70, 0.70:
bbox AP40:98.3601, 89.8058, 82.5364
bev  AP40:34.9062, 24.2195, 20.8772
3d   AP40:24.4504, 17.9385, 15.3246
aos  AP40:97.97, 89.21, 81.42
Car AP40@0.70, 0.50, 0.50:
bbox AP40:98.3601, 89.8058, 82.5364
bev  AP40:70.2495, 50.6545, 45.7686
3d   AP40:64.2237, 46.7250, 40.8438
aos  AP40:97.97, 89.21, 81.42

Overall AP40@easy, moderate, hard:
bbox AP40:76.5458, 60.7907, 54.8134
bev  AP40:14.7658, 10.1080, 8.5898
3d   AP40:10.4986, 7.4072, 6.5241
aos  AP40:70.57, 56.01, 50.29
-------------------------------------
[2024-05-15 06:07:38] Checkpoint is saved to 'checkpoints/epoch_200.pth'.
[2024-05-15 06:07:39] Checkpoint is saved to 'checkpoints/epoch_200_final.pth'.

五、测试

1、测试命令

### Evaluation
```bash
python test.py  --config_file       [FILL]      # Config file (.yaml file)
                --checkpoint_file   [FILL]      # Checkpoint file (.pth file)
                --gpu_id            [Optional]  # Index of GPU to use for testing (Default: 0)
                --evaluate                      # Perform evaluation (Quantitative Results)

Inference

python test.py  --config_file       [FILL]      # Config file (.yaml file)
                --checkpoint_file   [FILL]      # Checkpoint file (.pth file)
                --visualize                     # Perform visualization (Qualitative Results)
                --gpu_id            [Optional]  # Index of GPU to use for testing (Default: 0)
                --save_dir          [FILL]      # Path where visualization results will be saved to

python test.py --config_file /devdata/deeplearn/cv/3D/monocon-pytorch/config.yaml --checkpoint_file /devdata/deeplearn/cv/3D/monocon-pytorch/checkpoints/epoch_200_final.pth --visualize --gpu_id 0 --save_dir /devdata/deeplearn/cv/3D/monocon-pytorch/result --visualize

2、测试结果

2D 效果
3D 效果

bev效果