cutting edge;
Recently, there is a project in the laboratory about the target detection of tunnel spills. Regarding the work of labeling pictures, labeling needs to be used for labeling. I think it is very interesting. Let me record it.
1. To build the environment in the early stage, open cmd
----------------------初始查询------------------------------------ nvidia-smi #显示gpu当前信息状态 conda --version #检查Anaconda是否成功安装 nvcc -V #查看cuda版本 conda info –envs #检测目前安装了哪些环境 conda search --full-name python #检查目前有哪些版本的python可以安装 ----------------------创建环境------------------------------------- conda create -n learn python==3.9 #创建环境 python版本：3.9 环境名称：learn conda activate learn #激活环境 环境名称：learn
2. Download the labeling installation package in github
Download address: https://github.com/heartexlabs/labelImg
3. Modify the predefined_classes.txt in the data folder in the labelImg-master folder
Since only one category of spilled objects needs to be marked, only one can be kept
4. Unzip the installation package to the current folder, open the terminal in the labelImg-master folder to activate the environment
-----------激活先前创建的环境----------- conda activate learn ---------输入下面代码安装环境依赖------- conda install pyqt=5 conda install -c anaconda lxml pyrcc5 -o libs/resources.py resources.qrc --------------打开labeling-------------- python labelImg.py
5. Open labeling, set the format for saving labels, you can choose yolo, voc, etc.
Select the dataset directory to be labeled and the label storage directory
