When using Tensorflow 2 object detection api when do I normalize images of different sizes. Do I preprocess all the images to be same size then annotate with object bounding box? Or do the models somehow do the resizing internally and adjusts the predicted bounding box? The pre-trained models seems to have a preset sizes.

https://github.com/tensorflow/models/tree/master/research/object_detection/models https://github.com/tensorflow/models/tree/master/research/object_detection https://tensorflow-object-detection-api-tutorial.readthedocs.io/en/latest/index.html

CodePudding user response：

according to my understanding, it is not required to normalize or all the images do not need to be resized by yourself. API will do the job, all you need to do annotation.