mmdetection/projects/ViTDet/configs/lsj-100e_coco-instance.py

_base_ = [
    '../../../configs/_base_/default_runtime.py',
]

# dataset settings
dataset_type = 'CocoDataset'
data_root = 'data/coco/'
image_size = (1024, 1024)

backend_args = None

train_pipeline = [
    dict(type='LoadImageFromFile', backend_args=backend_args),
    dict(type='LoadAnnotations', with_bbox=True, with_mask=True),
    dict(type='RandomFlip', prob=0.5),
    dict(
        type='RandomResize',
        scale=image_size,
        ratio_range=(0.1, 2.0),
        keep_ratio=True),
    dict(
        type='RandomCrop',
        crop_type='absolute_range',
        crop_size=image_size,
        recompute_bbox=True,
        allow_negative_crop=True),
    dict(type='FilterAnnotations', min_gt_bbox_wh=(1e-2, 1e-2)),
    dict(type='Pad', size=image_size, pad_val=dict(img=(114, 114, 114))),
    dict(type='PackDetInputs')
]

test_pipeline = [
    dict(type='LoadImageFromFile', backend_args=backend_args),
    dict(type='Resize', scale=image_size, keep_ratio=True),
    dict(type='Pad', size=image_size, pad_val=dict(img=(114, 114, 114))),
    dict(type='LoadAnnotations', with_bbox=True, with_mask=True),
    dict(
        type='PackDetInputs',
        meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape',
                   'scale_factor'))
]

train_dataloader = dict(
    batch_size=4,
    num_workers=8,
    persistent_workers=True,
    sampler=dict(type='DefaultSampler', shuffle=True),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        ann_file='annotations/instances_train2017.json',
        data_prefix=dict(img='train2017/'),
        filter_cfg=dict(filter_empty_gt=True, min_size=32),
        pipeline=train_pipeline))

val_dataloader = dict(
    batch_size=1,
    num_workers=2,
    persistent_workers=True,
    drop_last=False,
    sampler=dict(type='DefaultSampler', shuffle=False),
    dataset=dict(
        type=dataset_type,
        data_root=data_root,
        ann_file='annotations/instances_val2017.json',
        data_prefix=dict(img='val2017/'),
        test_mode=True,
        pipeline=test_pipeline))
test_dataloader = val_dataloader

val_evaluator = dict(
    type='CocoMetric',
    ann_file=data_root + 'annotations/instances_val2017.json',
    metric=['bbox', 'segm'],
    format_only=False)
test_evaluator = val_evaluator

optim_wrapper = dict(
    type='AmpOptimWrapper',
    constructor='LayerDecayOptimizerConstructor',
    paramwise_cfg={
        'decay_rate': 0.7,
        'decay_type': 'layer_wise',
        'num_layers': 12,
    },
    optimizer=dict(
        type='AdamW',
        lr=0.0001,
        betas=(0.9, 0.999),
        weight_decay=0.1,
    ))

# 100 ep = 184375 iters * 64 images/iter / 118000 images/ep
max_iters = 184375
interval = 5000
dynamic_intervals = [(max_iters // interval * interval + 1, max_iters)]
param_scheduler = [
    dict(
        type='LinearLR', start_factor=0.001, by_epoch=False, begin=0, end=250),
    dict(
        type='MultiStepLR',
        begin=0,
        end=max_iters,
        by_epoch=False,
        # 88 ep = [163889 iters * 64 images/iter / 118000 images/ep
        # 96 ep = [177546 iters * 64 images/iter / 118000 images/ep
        milestones=[163889, 177546],
        gamma=0.1)
]

train_cfg = dict(
    type='IterBasedTrainLoop',
    max_iters=max_iters,
    val_interval=interval,
    dynamic_intervals=dynamic_intervals)
val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')

default_hooks = dict(
    logger=dict(type='LoggerHook', interval=50),
    checkpoint=dict(
        type='CheckpointHook',
        by_epoch=False,
        save_last=True,
        interval=interval,
        max_keep_ckpts=5))
vis_backends = [
    dict(type='LocalVisBackend'),
    dict(type='TensorboardVisBackend')
]
visualizer = dict(
    type='DetLocalVisualizer', vis_backends=vis_backends, name='visualizer')
log_processor = dict(type='LogProcessor', window_size=50, by_epoch=False)

auto_scale_lr = dict(base_batch_size=64)