frustum-pointnets復現過程+遇到的問題+解決方法

frustum-pointnets復現過程+遇到的問題+解決方法

系統Ubuntu16.04
1.下載KITTI數據集：

並按照README中的格式解壓並重組數據集：

2.運行環境配置：
在復現KPConv的環境（‘python3’）下繼續，首先安裝一些包，已經安裝的包不再安裝（可KPConv復現鏈接中查看）：

conda install pillow
pip install opencv-python

3.數據準備
在對應目錄終端下，運行sh scripts/command_prep_data.sh

4.修改部分代碼
（1）frustum-pointnets/train/provider.py

import cPickle as pickle    ---->    import _pickle as pickle

將self.id_list = pickle.load(fp)等涉及pickle.load的函數，修改爲

self.id_list = pickle.load(fp, encoding='latin1')    # or “ encoding='iso-8859-1' ”

（2）frustum-pointnets/kitti/kitti_util.py
將cv2.CV_AA修改爲cv2.LINE_AA（因爲opencv-python==4.1.0.25）
（3）frustum-pointnets/kitti/prepare_data.py

 import cPickle as pickle    ---->    import _pickle as pickle

（4）遇到問題：TypeError: 'float' object cannot be interpreted as an integer
解決辦法：frustum-pointnets/train/train.py

train_one_epoch函數和eval_one_epoch函數中：
num_batches = len(TRAIN_DATASET)/BATCH_SIZE
---->
num_batches = len(TRAIN_DATASET)//BATCH_SIZE

5.運行CUDA_VISIBLE_DEVICES=0 sh scripts/command_train_v1.sh

6.修改部分代碼
（1）遇到問題：TabError: inconsistent use of tabs and spaces in indentation
解決辦法：frustum-pointnets/train/test.py
修改line 231和line 310中的batch_output, batch_center_pred, \語句，將該語句前面的tab換成空格。
（2）frustum-pointnets/train/test.py

 import cPickle as pickle    ---->    import _pickle as pickle

（3）frustum-pointnets/train/test.py

在 def inference(sess, ops, pc, one_hot_vec, batch_size)中：
num_batches = pc.shape[0]/batch_size
---->
num_batches = pc.shape[0]//batch_size

（4）需要重新編譯frustum-pointnets/train/kitti_eval/evaluate_object_3d_offline：
g++ evaluate_object_3d_offline.cpp
遇到問題：
fatal error: boost/numeric/ublas/matrix.hpp: No such file or directory
解決辦法：

sudo apt-get update
sudo apt-get install libboost-all-dev

（5）遇到問題：

sh: 1: gnuplot: not found

sh: 1: pdfcrop: not found

解決辦法：

sudo apt-get install libx11-dev
sudo apt-get install gnuplot

sudo apt install texlive-extra-utils

7.運行CUDA_VISIBLE_DEVICES=0 sh scripts/command_test_v1.sh

8.修改部分代碼
（1）遇到問題：

tf_sampling_compile.sh: 2: tf_sampling_compile.sh: /usr/local/cuda-8.0/bin/nvcc: not found
g++: error: tf_sampling_g.cu.o: No such file or directory

解決辦法：
在frustum-pointnets/model/tf_ops/sampling/下打開終端：
conda activate python3

（2）

9.運行CUDA_VISIBLE_DEVICES=0 sh scripts/command_test_v2.sh

10.git上下載frustum-pointnets-view-results，然後修改view_results.py中：
找到line 32、line 40、line 41、line 213、line 216的語句，將該語句前面的tab換成空格，然後運行python kitti/view_results.py

未完待續…