Help!!!【Mask2former】All predictions are colored by full colors in visualization!
Opened this issue · 1 comments
FreeWillThorn commented
I am doing instance segmentation, and Mask2former performs great, aAcc reachs almost 99% which is unbelievable.
When i use visualization=dict(type='SegVisualizationHook',draw=True, interval=10)) to visualize the result, the predictions are fully colored instead of the target area.
How could the wrong prediction gives such wonderful class? Do you guys know how to fix it?
-
Dataset is valid: Performs normally on the the other methods that contained in MMsegmentation
-
Config file: mask2former_swin-l-in22k-384x384-pre_8xb2-160k_ade20k-640x640.py
Config file1 :
_base_ = [
'../_base_/default_runtime.py', '../_base_/datasets/ade20k_640x640.py'
]
pretrained = 'https://download.openmmlab.com/mmsegmentation/v0.5/pretrain/swin/swin_base_patch4_window12_384_20220317-55b0104a.pth' # noqa
crop_size = (640, 640)
data_preprocessor = dict(
type='SegDataPreProcessor',
mean=[123.675, 116.28, 103.53],
std=[58.395, 57.12, 57.375],
bgr_to_rgb=True,
pad_val=0,
seg_pad_val=255,
size=crop_size)
num_classes = 7
depths = [2, 2, 18, 2]
model = dict(
type='EncoderDecoder',
data_preprocessor=data_preprocessor,
backbone=dict(
type='SwinTransformer',
pretrain_img_size=384,
embed_dims=128,
depths=depths,
num_heads=[4, 8, 16, 32],
window_size=12,
mlp_ratio=4,
qkv_bias=True,
qk_scale=None,
drop_rate=0.,
attn_drop_rate=0.,
drop_path_rate=0.3,
patch_norm=True,
out_indices=(0, 1, 2, 3),
with_cp=False,
frozen_stages=-1,
init_cfg=dict(type='Pretrained', checkpoint=pretrained)),
decode_head=dict(
type='Mask2FormerHead',
in_channels=[128, 256, 512, 1024],
strides=[4, 8, 16, 32],
feat_channels=256,
out_channels=7,
num_classes=num_classes,
num_queries=100,
num_transformer_feat_level=3,
align_corners=False,
pixel_decoder=dict(
type='mmdet.MSDeformAttnPixelDecoder',
num_outs=3,
norm_cfg=dict(type='GN', num_groups=32),
act_cfg=dict(type='ReLU'),
encoder=dict( # DeformableDetrTransformerEncoder
num_layers=6,
layer_cfg=dict( # DeformableDetrTransformerEncoderLayer
self_attn_cfg=dict( # MultiScaleDeformableAttention
embed_dims=256,
num_heads=8,
num_levels=3,
num_points=4,
im2col_step=64,
dropout=0.0,
batch_first=True,
norm_cfg=None,
init_cfg=None),
ffn_cfg=dict(
embed_dims=256,
feedforward_channels=1024,
num_fcs=2,
ffn_drop=0.0,
act_cfg=dict(type='ReLU', inplace=True))),
init_cfg=None),
positional_encoding=dict( # SinePositionalEncoding
num_feats=128, normalize=True),
init_cfg=None),
enforce_decoder_input_project=False,
positional_encoding=dict( # SinePositionalEncoding
num_feats=128, normalize=True),
transformer_decoder=dict( # Mask2FormerTransformerDecoder
return_intermediate=True,
num_layers=9,
layer_cfg=dict( # Mask2FormerTransformerDecoderLayer
self_attn_cfg=dict( # MultiheadAttention
embed_dims=256,
num_heads=8,
attn_drop=0.0,
proj_drop=0.0,
dropout_layer=None,
batch_first=True),
cross_attn_cfg=dict( # MultiheadAttention
embed_dims=256,
num_heads=8,
attn_drop=0.0,
proj_drop=0.0,
dropout_layer=None,
batch_first=True),
ffn_cfg=dict(
embed_dims=256,
feedforward_channels=2048,
num_fcs=2,
act_cfg=dict(type='ReLU', inplace=True),
ffn_drop=0.0,
dropout_layer=None,
add_identity=True)),
init_cfg=None),
loss_cls=dict(
type='mmdet.CrossEntropyLoss',
use_sigmoid=False,
loss_weight=2.0,
reduction='mean',
class_weight=[1.0] * num_classes + [0.1]),
loss_mask=dict(
type='mmdet.CrossEntropyLoss',
use_sigmoid=True,
reduction='mean',
loss_weight=5.0),
loss_dice=dict(
type='mmdet.DiceLoss',
use_sigmoid=True,
activate=True,
reduction='mean',
naive_dice=True,
eps=1.0,
loss_weight=5.0),
train_cfg=dict(
num_points=12544,
oversample_ratio=3.0,
importance_sample_ratio=0.75,
assigner=dict(
type='mmdet.HungarianAssigner',
match_costs=[
dict(type='mmdet.ClassificationCost', weight=2.0),
dict(
type='mmdet.CrossEntropyLossCost',
weight=5.0,
use_sigmoid=True),
dict(
type='mmdet.DiceCost',
weight=5.0,
pred_act=True,
eps=1.0)
]),
sampler=dict(type='mmdet.MaskPseudoSampler'))),
train_cfg=dict(),
test_cfg=dict(mode='whole'))
# dataset config
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='LoadAnnotations', reduce_zero_label=True),
dict(
type='RandomChoiceResize',
scales=[int(x * 0.1 * 640) for x in range(5, 21)],
resize_type='ResizeShortestEdge',
max_size=2560),
dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
dict(type='RandomFlip', prob=0.5),
dict(type='PhotoMetricDistortion'),
dict(type='PackSegInputs')
]
train_dataloader = dict(batch_size=4, dataset=dict(pipeline=train_pipeline))
# set all layers in backbone to lr_mult=0.1
# set all norm layers, position_embeding,
# query_embeding, level_embeding to decay_multi=0.0
backbone_norm_multi = dict(lr_mult=0.1, decay_mult=0.0)
backbone_embed_multi = dict(lr_mult=0.1, decay_mult=0.0)
embed_multi = dict(lr_mult=1.0, decay_mult=0.0)
custom_keys = {
'backbone': dict(lr_mult=0.1, decay_mult=1.0),
'backbone.patch_embed.norm': backbone_norm_multi,
'backbone.norm': backbone_norm_multi,
'absolute_pos_embed': backbone_embed_multi,
'relative_position_bias_table': backbone_embed_multi,
'query_embed': embed_multi,
'query_feat': embed_multi,
'level_embed': embed_multi
}
custom_keys.update({
f'backbone.stages.{stage_id}.blocks.{block_id}.norm': backbone_norm_multi
for stage_id, num_blocks in enumerate(depths)
for block_id in range(num_blocks)
})
custom_keys.update({
f'backbone.stages.{stage_id}.downsample.norm': backbone_norm_multi
for stage_id in range(len(depths) - 1)
})
# optimizer
optimizer = dict(
type='AdamW', lr=0.0001, weight_decay=0.05, eps=1e-8, betas=(0.9, 0.999))
optim_wrapper = dict(
type='OptimWrapper',
optimizer=optimizer,
clip_grad=dict(max_norm=0.01, norm_type=2),
paramwise_cfg=dict(custom_keys=custom_keys, norm_decay_mult=0.0))
# learning policy
param_scheduler = [
dict(
type='PolyLR',
eta_min=0,
power=0.9,
begin=0,
end=160000,
by_epoch=False)
]
# training schedule for 160k
train_cfg = dict(
type='IterBasedTrainLoop', max_iters=160000, val_interval=5000)
val_cfg = dict(type='ValLoop')
test_cfg = dict(type='TestLoop')
default_hooks = dict(
timer=dict(type='IterTimerHook'),
logger=dict(type='LoggerHook', interval=50, log_metric_by_epoch=False),
param_scheduler=dict(type='ParamSchedulerHook'),
checkpoint=dict(
type='CheckpointHook', by_epoch=False, interval=5000,
save_best='mIoU'),
sampler_seed=dict(type='DistSamplerSeedHook'),
visualization=dict(type='SegVisualizationHook',draw=True, interval=10))
# Default setting for scaling LR automatically
# - `enable` means enable scaling LR automatically
# or not by default.
# - `base_batch_size` = (8 GPUs) x (2 samples per GPU).
auto_scale_lr = dict(enable=False, base_batch_size=16)
Config file2 :
_base_ = ['./mask2former_swin-b-in1k-384x384-pre_8xb2-160k_ade20k-640x640.py']
pretrained = 'https://download.openmmlab.com/mmsegmentation/v0.5/pretrain/swin/swin_large_patch4_window12_384_22k_20220412-6580f57d.pth' # noqa
model = dict(
backbone=dict(
embed_dims=192,
num_heads=[6, 12, 24, 48],
init_cfg=dict(type='Pretrained', checkpoint=pretrained)),
decode_head=dict(num_queries=100, in_channels=[192, 384, 768, 1536]))
FreeWillThorn commented
