Hyper-parameter optimization in tensorflow object detection API

Question

Is there any way to specify hyper-parameter optimisation like Hyperopt or other in the config file of object detection API to fine tune the model?

Teetotalism

Answer 1

You can change following hyper parameters for object detection:

• learning rate
• momentum
• optimizer
• decay factor
• matched_threshold
• unmatched_threshold

Smallscale