> For the complete documentation index, see [llms.txt](https://pcl.gitbook.io/tutorial/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://pcl.gitbook.io/tutorial/part-3/part03-chapter04/part03-chapter04-voxelnet.md).

# VoxelNet (50%)

## 1. 개요

<https://adioshun.gitbooks.io/paper-3d-object-detection-and-tracking/content/2017-voxelnet-end-to-end-learning-for-point-cloud-based-3d-object-detection.html>

## 2. 설치 (Docker 기반)

### 데이터 준비

```
└── DATA_DIR
       ├── training   <-- training data
       |   ├── image_2
       |   ├── label_2
       |   └── velodyne
       └── validation  <--- evaluation data
       |   ├── image_2
       |   ├── label_2
       |   └── velodyne
```

### 도커 pull & 실행

```
$ docker pull adioshun/voxelnet
$ docker run --runtime=nvidia -it --privileged --network=host -v /tmp/.X11-unix:/tmp/.X11-unix --volume="$HOME/.Xauthority:/root/.Xauthority:rw" -e DISPLAY -v /media/{DATA_DIR}/datasets:/dataset --name 'voxelnet' adioshun/voxelnet /bin/bash
```

## 3. 실행

### 설정 수정(In the docker )

* `config.py`

```
# for dataset dir
__C.DATA_DIR = '/voxelnet/data/dataset'
__C.CALIB_DIR = '/voxelnet/data/dataset/training/calib'


# for gpu allocation
__C.GPU_AVAILABLE = '0,1'
__C.GPU_USE_COUNT = len(__C.GPU_AVAILABLE.split(','))
__C.GPU_MEMORY_FRACTION = 1
```

* `kitti_eval/launch_test.sh`

```
```

### train

```
python3 train.py --vis true
```

* log 저장 위치  : `log/default` #Tensorboard 지원&#x20;
* validation results : `predictions/{epoch number}/data`
* validation results(이미지) : `predictions/{epoch number}/vis` # --vis true 사용시 (기본 false)
* model 저장 위치 : `save_model/default`
* 학습된 model 저장 위치 :`save_model/pre_trained_car`

> Nvidia 1080 Ti GPUs로 약 3일이 소요 되므로 학습된 모델 사용을 권장 합니다.

학습 완료 후 **Learning Curve** 확인

```
python3 parse_log.py predictions
# predictions.jpg 생성
```

### Evaluate

```
$ python3 test.py -n default --vis True#학습된 결과물 활용 
$ python3 test.py -n pre_trained_car --vis True#사전 학습된 결과물 활용 `save_model/pre_trained_car`
```

* 결과 저장 폴더 : `predictions/data`
* 결과 저장 폴더(이미지) : `predictions/vis`# --vis true 사용시 (기본 false)

### 결과 확인

```
./kitti_eval/evaluate_object_3d_offline ./../data/dataset/validation/label_2/ ./prediction
```

## ![](https://i.imgur.com/L1UM4bO.png)

* [an unofficial inplementation of VoxelNet in TensorFlow](https://github.com/qianguih/voxelnet)


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://pcl.gitbook.io/tutorial/part-3/part03-chapter04/part03-chapter04-voxelnet.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.