Search code examples
pythontensorflowobject-detection

Changing the anchor size for transfer learning

I want to detect some small objects and I wonder if changing anchor size in Faster_rcnn.config file will make the model perform better in transfer learning or the model will require to be trained from scratch? I am using pre-trained models on the Object detection model zoo as the fine-tuning checkpoint.

The ".config" file has the following parameters:

first_stage_anchor_generator {
grid_anchor_generator {
scales: [0.25, 0.5, 1.0, 2.0]
aspect_ratios: [0.5, 1.0, 2.0]
height_stride: 8
width_stride: 8

Thanks for your answers and help!

Solution

  • The best way around is by using Kmeans for clustering. This is a very useful document. Kmeans