panformer_pvtb5_24e_coco_panoptic.py

_base_ = './base.py'

model = dict(
    # get pvt_v2_b5_22k 
    # wget /~https://github.com/whai362/PVT/releases/download/v2/pvt_v2_b5_22k.pth
    pretrained='./checkpoints/pvt_v2_b5_22k.pth',
    backbone=dict(
        type='pvt_v2_b5',
        out_indices=(1, 2, 3),
       ),
    neck=dict(
        type='ChannelMapper',
        in_channels=[128, 320, 512], 
     ),
    bbox_head=dict(
        quality_threshold_things=0.3,
        quality_threshold_stuff=0.3,
    )
)