mask_rcnn_x101_32x4d_fpn_mstrain-poly_3x_coco.py

_base_ = [
    '../common/mstrain-poly_3x_coco_instance.py',
    '../_base_/models/mask_rcnn_r50_fpn.py'
]

model = dict(
    backbone=dict(
        type='ResNeXt',
        depth=101,
        groups=32,
        base_width=4,
        num_stages=4,
        out_indices=(0, 1, 2, 3),
        frozen_stages=1,
        norm_cfg=dict(type='BN', requires_grad=True),
        style='pytorch',
        init_cfg=dict(
            type='Pretrained', checkpoint='open-mmlab://resnext101_32x4d')))