Why Average Precision and Average Recall = -1 when using tensorflow to do object detection?

Question

I'm using Google Colab, Tensorflow version 1, SSD mobile net v2 CoCo model to train a custom dataset. But evaluation on a custom dataset above the image that AP and AR = -1. I don't understand why the result appears?

Can you answer this question? Thank you.

Answer 1

For any object detection model using tensorflow the Average Precision and Average Recall value can be 0 but not negative.

If it is coming -1 then the reason may be as follows:

• It may be a condition of underfitting, you should get more training data. You should increase the size or number of parameters in the model. Try to add new neurons in existing layers or training the model a little bit longer until the cost function is minimized.

• Your model may not be able to distinguish the image so that do add similar images to your dataset.

• Check the format of your inputs, model might not be getting the inputs in proper format.