This post is a complementary part for Building Your Own Real-Time Object Detection App: Roboflow(YOLOv8) and Streamlit. Will guide us in how to label images with another tool just in case we are going to use the images for a TensorFlow notebook.
If you want to use a tool that doesn’t require WiFi I recommend to you labelimg is an open source tool that you can find in Github. I had problems installing it using a virtual environment and Conda so I recommend you go to this link and download the binary file to use directly in Windows. It may give you some warning but click in More info and then Run Anyway.
Before diving in to Roboflow I’ll give you some tips in case you want to label them in labelimg:
There is a predefined classes txt file so if you want to save time and insted of typing everytime that you label a image in labelimg you can modify this document and add the classes that you going to work with, just open and add the classes to the file.
Now open the labelimg and click in Open Dir to see your files in the tool bar in the right. Just click Select Folder when you have the folder with the images.
To save time click W and it will appear the select option where yoy create a label square over the object you wan to classify.
Use the shorcut ctrl+s to save the xml file. And then you can click d to go to the next image with out labels.
Make sure each image has the same name for it xml label file this is an important step.
After having a folder with all of this. You can upload to Google Colab or work with it locally.
In this part of tutorial series, we have introduced you to the labelimg. This post give you a small guide on how to do it.