mmfewshot
/
configs
/
detection
/
fsce
/
coco
/
fsce_r101_fpn_coco_base-training.py

_base_ = [
    '../../_base_/datasets/fine_tune_based/base_coco.py',
    '../../_base_/schedules/schedule.py',
    '../../_base_/models/faster_rcnn_r50_caffe_fpn.py',
    '../../_base_/default_runtime.py'
]
lr_config = dict(warmup_iters=1000, step=[85000, 100000])
runner = dict(max_iters=110000)
# model settings
model = dict(
    pretrained='open-mmlab://detectron2/resnet101_caffe',
    backbone=dict(depth=101, style='caffe'),
    roi_head=dict(bbox_head=dict(num_classes=60)))