I am trying to detect multiple objects such as road signs, pedestrians etc using Tensorflow object detection. However I am unsure whether to have a single model for each object type or have a single model which can detect multiple object types. How does performance in terms of accuracy, resource consumption etc. vary with each approach? Are there any other factors to consider?
1 Answers
Single model which can detect multiple object types is a better choice, because the speed will be N times faster than N models, memory usage will also be reduced by N times, and the accuracy will be about the same.
Related