# Basic Settings
default_scope = 'mmdet'
log_level = 'INFO'
load_from = 'checkpoints/yolox_x_Black_only.pth' # Changed to YOLOX-X checkpoint
resume = False
# Environment Configuration
env_cfg = dict(
cudnn_benchmark=False,
dist_cfg=dict(backend='nccl'),
mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0)
)
# Dataset and Data Root Configuration
data_root = 'data/White-lined-designs-Test-COCO/'
dataset_type = 'CocoDataset'
class_name = ['stethoscope']
metainfo = dict(
classes=class_name,
palette=[(20, 220, 60)]
)
# Image Size Configuration
img_scale = (640, 480)
img_scales = [(640, 480)]
# Model Configuration
model = dict(
type='YOLOX',
data_preprocessor=dict(
type='DetDataPreprocessor',
pad_size_divisor=32,
batch_augments=[
dict(
type='BatchSyncRandomResize',
random_size_range=(480, 480),
size_divisor=32,
interval=10
),
],
),
backbone=dict(
type='CSPDarknet',
deepen_factor=1.33, # Increased from 1.0 for YOLOX-X
widen_factor=1.25, # Increased from 1.0 for YOLOX-X
out_indices=(2, 3, 4),
use_depthwise=False,
spp_kernal_sizes=(5, 9, 13),
norm_cfg=dict(type='BN', momentum=0.03, eps=0.001),
act_cfg=dict(type='Swish'),
),
neck=dict(
type='YOLOXPAFPN',
in_channels=[320, 640, 1280], # Changed according to widen_factor
out_channels=320, # Changed according to widen_factor
num_csp_blocks=4, # Increased from 3 for YOLOX-X
use_depthwise=False,
upsample_cfg=dict(scale_factor=2, mode='nearest'),
norm_cfg=dict(type='BN', momentum=0.03, eps=0.001),
act_cfg=dict(type='Swish'),
),
bbox_head=dict(
type='YOLOXHead',
num_classes=1,
in_channels=320, # Changed according to neck out_channels
feat_channels=320, # Changed according to neck out_channels
strides=(8, 16, 32),
stacked_convs=2,
use_depthwise=False,
norm_cfg=dict(type='BN', momentum=0.03, eps=0.001),
act_cfg=dict(type='Swish'),
loss_cls=dict(
type='CrossEntropyLoss',
use_sigmoid=True,
reduction='sum',
loss_weight=1.0
),
loss_bbox=dict(
type='IoULoss',
mode='square',
eps=1e-16,
reduction='sum',
loss_weight=5.0
),
loss_obj=dict(
type='CrossEntropyLoss',
use_sigmoid=True,
reduction='sum',
loss_weight=1.0
),
loss_l1=dict(type='L1Loss', reduction='sum', loss_weight=1.0),
),
train_cfg=dict(assigner=dict(type='SimOTAAssigner', center_radius=2.5)),
test_cfg=dict(score_thr=0.01, nms=dict(type='nms', iou_threshold=0.65))
)
# Training Configuration
max_epochs = 300
num_last_epochs = 15
base_lr = 0.01 * 64 / 128 # Adjusted for batch size
interval = 10
# Optimization Configuration
optim_wrapper = dict(
type='OptimWrapper',
optimizer=dict(
type='SGD',
lr=0.01 * 64 / 128, # Adjusted for batch size
momentum=0.9,
weight_decay=0.0005,
nesterov=True
),
paramwise_cfg=dict(
norm_decay_mult=0.0,
bias_decay_mult=0.0
)
)
# Learning Rate Schedule
param_scheduler = [
dict(
type='mmdet.QuadraticWarmupLR',
by_epoch=True,
begin=0,
end=5,
convert_to_iter_based=True
),
dict(
type='CosineAnnealingLR',
T_max=285,
eta_min=0.0005 * 64 / 128, # Adjusted for batch size
begin=5,
end=285,
by_epoch=True,
convert_to_iter_based=True
),
dict(
type='ConstantLR',
factor=1,
begin=285,
end=300,
by_epoch=True
),
]
# Custom Hooks Configuration
custom_hooks = [
dict(
type='YOLOXModeSwitchHook',
num_last_epochs=15,
priority=48
),
dict(
type='SyncNormHook',
priority=48
),
dict(
type='EMAHook',
ema_type='ExpMomentumEMA',
momentum=0.0001,
update_buffers=True,
priority=49
),
]
# Default Hooks Configuration
default_hooks = dict(
timer=dict(type='IterTimerHook'),
logger=dict(type='LoggerHook', interval=50),
param_scheduler=dict(type='ParamSchedulerHook'),
checkpoint=dict(type='CheckpointHook', interval=10, max_keep_ckpts=3),
sampler_seed=dict(type='DistSamplerSeedHook'),
visualization=dict(type='DetVisualizationHook')
)
# Data Loading Configuration
train_dataloader = dict(
batch_size=8, # Might need to reduce depending on GPU memory
num_workers=4,
persistent_workers=True,
sampler=dict(type='DefaultSampler', shuffle=True),
dataset=dict(
type='MultiImageMixDataset',
dataset=dict(
type='CocoDataset',
data_root=data_root,
ann_file='train/annotations/instances_default.json',
data_prefix=dict(img='train/images/'),
metainfo=metainfo,
filter_cfg=dict(filter_empty_gt=False, min_size=32),
pipeline=[
dict(type='LoadImageFromFile', backend_args=None),
dict(type='LoadAnnotations', with_bbox=True),
],
backend_args=None
),
pipeline=[
dict(type='Mosaic', img_scale=(640, 480), pad_val=114.0),
dict(
type='RandomAffine',
scaling_ratio_range=(0.1, 2),
border=(-320, -240)
),
dict(
type='MixUp',
img_scale=(640, 480),
ratio_range=(0.8, 1.6),
pad_val=114.0
),
dict(type='YOLOXHSVRandomAug'),
dict(type='RandomFlip', prob=0.5),
dict(type='Resize', scale=(640, 480), keep_ratio=True),
dict(
type='Pad',
pad_to_square=True,
pad_val=dict(img=(114.0, 114.0, 114.0))
),
dict(
type='FilterAnnotations',
min_gt_bbox_wh=(1, 1),
keep_empty=False
),
dict(type='PackDetInputs')
]
)
)
# Validation Configuration
val_dataloader = dict(
batch_size=8, # Might need to reduce depending on GPU memory
num_workers=4,
persistent_workers=True,
drop_last=False,
sampler=dict(type='DefaultSampler', shuffle=False),
dataset=dict(
type='CocoDataset',
data_root=data_root,
ann_file='valid/annotations/instances_default.json',
data_prefix=dict(img='valid/images/'),
test_mode=True,
metainfo=metainfo,
pipeline=[
dict(type='LoadImageFromFile', backend_args=None),
dict(type='Resize', scale=(640, 480), keep_ratio=True),
dict(
type='Pad',
pad_to_square=True,
pad_val=dict(img=(114.0, 114.0, 114.0))
),
dict(type='LoadAnnotations', with_bbox=True),
dict(
type='PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape', 'scale_factor')
)
]
)
)
# Test Configuration
test_dataloader = val_dataloader
# Metrics and Evaluation Configuration
val_evaluator = dict(
type='CocoMetric',
ann_file=data_root + 'valid/annotations/instances_default.json',
metric='bbox',
backend_args=None
)
test_evaluator = val_evaluator
# Training and Testing Loops Configuration
train_cfg = dict(
type='EpochBasedTrainLoop',
max_epochs=max_epochs,
val_interval=10
)
val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')
# Test Time Augmentation Configuration
tta_model = dict(
type='DetTTAModel',
tta_cfg=dict(nms=dict(type='nms', iou_threshold=0.65), max_per_img=100)
)
tta_pipeline = [
dict(type='LoadImageFromFile', backend_args=None),
dict(
type='TestTimeAug',
transforms=[
[
dict(type='Resize', scale=(640, 480), keep_ratio=True),
dict(type='Resize', scale=(320, 240), keep_ratio=True),
dict(type='Resize', scale=(960, 720), keep_ratio=True),
],
[
dict(type='RandomFlip', prob=1.0),
dict(type='RandomFlip', prob=0.0),
],
[
dict(type='Pad', pad_to_square=True, pad_val=dict(img=(114.0, 114.0, 114.0))),
],
[
dict(type='LoadAnnotations', with_bbox=True),
],
[
dict(
type='PackDetInputs',
meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape', 'scale_factor', 'flip', 'flip_direction')
),
]
]
)
]
# Visualization Configuration
vis_backends = [dict(type='LocalVisBackend')]
visualizer = dict(
type='DetLocalVisualizer',
vis_backends=vis_backends,
name='visualizer'
)
# Auto Scale Learning Rate Configuration
auto_scale_lr = dict(enable=False, base_batch_size=64)
log_processor = dict(type='LogProcessor', window_size=50, by_epoch=True)
