mmdetection tips (strongly recommended to gradually improve)

• Tip 1: All configuration files in mmdetection have such a line of code. If target detection is to be done, reference coco_detection.py, and if target detection is to be done, reference coco_instance.py. When training Swin Transformer, the modification here is particularly common, depending on what you need to do, here we are doing target detection.

  _base_ = [
      '../_base_/models/mask_rcnn_r50_fpn.py',
      '../_base_/datasets/coco_detection.py', 
      '../_base_/schedules/schedule_1x.py', '../_base_/default_runtime.py'
  ]

• Tip 2: In coco_detection.py, there is a parameter data_root, which is the path to store our dataset.

  dataset_type = 'CocoDataset'
  data_root = 'data/coco/'
  img_norm_cfg = dict(mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)