configs/other_net_cofigs/dino-5scale_swin-l-12e_hsi4x.py

_base_ = [
    './datasets/hsi_detection4x.py', '../_base_/default_runtime.py'
    # './datasets/hsi_detection.py', '../_base_/default_runtime.py'
]

fp16 = dict(loss_scale=512.)
# pretrained = 'https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_large_patch4_window12_384_22k.pth'  # noqa

pretrained = 'https://download.openmmlab.com/mmclassification/v0/swin-transformer/convert/swin_large_patch4_window12_384_22kto1k-0a40944b.pth'
num_levels = 5
in_channels = 30

model = dict(
    type='DINO',
    num_queries=900,  # num_matching_queries 900
    with_box_refine=True,
    as_two_stage=True,
    num_feature_levels=num_levels,
    data_preprocessor=dict(
        type='HSIDetDataPreprocessor',
        pad_size_divisor=1),
    backbone=dict(
        type='SwinTransformer',
        pretrain_img_size=384,
        in_channels=in_channels,
        embed_dims=192,
        depths=[2, 2, 18, 2],
        num_heads=[6, 12, 24, 48],
        window_size=12,
        mlp_ratio=4,
        qkv_bias=True,
        qk_scale=None,
        drop_rate=0.,
        attn_drop_rate=0.,
        drop_path_rate=0.2,
        patch_norm=True,
        out_indices=(0, 1, 2, 3),
        # Please only add indices that would be used
        # in FPN, otherwise some parameter will not be used
        with_cp=True,
        convert_weights=True,
        init_cfg=dict(type='Pretrained', checkpoint=pretrained)
    ),
    neck=dict(
        type='ChannelMapper',
        in_channels=[192, 384, 768, 1536],
        kernel_size=1,
        out_channels=256,
        act_cfg=None,
        norm_cfg=dict(type='GN', num_groups=32),
        num_outs=num_levels),
    encoder=dict(
        num_layers=6,
        layer_cfg=dict(
            self_attn_cfg=dict(embed_dims=256, num_levels=num_levels,
                               dropout=0.0),  # 0.1 for DeformDETR
            ffn_cfg=dict(
                embed_dims=256,
                feedforward_channels=2048,  # 1024 for DeformDETR
                ffn_drop=0.0))),  # 0.1 for DeformDETR
    decoder=dict(
        num_layers=6,
        return_intermediate=True,
        layer_cfg=dict(
            self_attn_cfg=dict(embed_dims=256, num_heads=8,
                               dropout=0.0),  # 0.1 for DeformDETR
            cross_attn_cfg=dict(embed_dims=256, num_levels=num_levels,
                                dropout=0.0),  # 0.1 for DeformDETR
            ffn_cfg=dict(
                embed_dims=256,
                feedforward_channels=2048,  # 1024 for DeformDETR
                ffn_drop=0.0)),  # 0.1 for DeformDETR
        post_norm_cfg=None),
    positional_encoding=dict(
        num_feats=128,
        normalize=True,
        offset=0.0,  # -0.5 for DeformDETR
        temperature=20),  # 10000 for DeformDETR
    bbox_head=dict(
        type='DINOHead',
        num_classes=8,
        sync_cls_avg_factor=True,
        loss_cls=dict(
            type='FocalLoss',
            use_sigmoid=True,
            gamma=2.0,
            alpha=0.25,
            loss_weight=1.0),  # 2.0 in DeformDETR
        loss_bbox=dict(type='L1Loss', loss_weight=5.0),
        loss_iou=dict(type='GIoULoss', loss_weight=2.0)),
    dn_cfg=dict(  # TODO: Move to model.train_cfg ?
        label_noise_scale=0.5,
        box_noise_scale=1.0,  # 0.4 for DN-DETR
        group_cfg=dict(dynamic=True, num_groups=None,
                       num_dn_queries=100)),  # TODO: half num_dn_queries
    # training and testing settings
    train_cfg=dict(
        assigner=dict(
            type='HungarianAssigner',
            match_costs=[
                dict(type='FocalLossCost', weight=2.0),
                dict(type='BBoxL1Cost', weight=5.0, box_format='xywh'),
                dict(type='IoUCost', iou_mode='giou', weight=2.0)
            ])),
    test_cfg=dict(max_per_img=300))  # 100 for DeformDETR


# optimizer
optim_wrapper = dict(
    type='OptimWrapper',
    optimizer=dict(
        type='AdamW',
        lr=0.0001,  # 0.0002 for DeformDETR
        weight_decay=0.0001),
    clip_grad=dict(max_norm=0.1, norm_type=2),
    paramwise_cfg=dict(custom_keys={'backbone': dict(lr_mult=0.1)})
)  # custom_keys contains sampling_offsets and reference_points in DeformDETR  # noqa

# learning policy
max_epochs = 12
train_cfg = dict(
    type='EpochBasedTrainLoop', max_epochs=max_epochs, val_interval=12)

val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')

param_scheduler = [
    dict(
        type='MultiStepLR',
        begin=0,
        end=max_epochs,
        by_epoch=True,
        milestones=[11],
        gamma=0.1)
]



train_dataloader = dict(
    batch_size=4,)
test_dataloader = dict(
    batch_size=1,)

# NOTE: `auto_scale_lr` is for automatically scaling LR,
# USER SHOULD NOT CHANGE ITS VALUES.
# base_batch_size = (8 GPUs) x (2 samples per GPU)
auto_scale_lr = dict(base_batch_size=4)