How should I adjust the training parameters, such as base_anchor_shapes? #1872

Activity

1. fanhenghui

   changed the title [-]I am currently researching the detection sub-project. When training for lung nodules smaller than 3mm, how should I adjust the training parameters, such as base_anchor_shapes?[/-] [+]How should I adjust the training parameters, such as base_anchor_shapes?[/+] on Nov 6, 2024
2. KumoLiu commented on Nov 6, 2024

   

   KumoLiu

   on Nov 6, 2024

   Contributor

   More actions

   Hi @fanhenghui, adjust base_anchor_shapes to include smaller anchor box sizes that correspond to the pixel scale of 3mm nodules in your images. If your image resolution makes 3mm correspond to, say, 5-7 pixels, set anchors to sizes close to these dimensions.

   tutorials/detection/luna16_training.py

   Lines 164 to 165 in fc15cdd

   	# base_anchor_shapes: anchor shape for the most high-resolution output,
   	# when target boxes are small, set it smaller
3. Project-MONAI

   locked and limited conversation to collaborators on Nov 6, 2024
4. KumoLiu

   converted this issue into a discussion #1873 on Nov 6, 2024