yolov5_s-v61_syncbn_fast_1xb4-300e_balloon.py

_base_ = './yolov5_s-v61_syncbn_fast_8xb16-300e_coco.py'

data_root = 'data/balloon/'

train_batch_size_per_gpu = 4
train_num_workers = 2

metainfo = {
    'CLASSES': ('balloon', ),
    'PALETTE': [
        (220, 20, 60),
    ]
}

train_dataloader = dict(
    batch_size=train_batch_size_per_gpu,
    num_workers=train_num_workers,
    dataset=dict(
        data_root=data_root,
        metainfo=metainfo,
        data_prefix=dict(img='train/'),
        ann_file='train.json'))

val_dataloader = dict(
    dataset=dict(
        data_root=data_root,
        metainfo=metainfo,
        data_prefix=dict(img='val/'),
        ann_file='val.json'))

test_dataloader = val_dataloader

val_evaluator = dict(ann_file=data_root + 'val.json')

test_evaluator = val_evaluator

model = dict(bbox_head=dict(head_module=dict(num_classes=1)))

default_hooks = dict(logger=dict(interval=1))