Freeze the layer in object detection

vision
1

if i want to do captioning and have a model doing object detection and will train it on my data.
Should i freeze a layer of the model