Results isssue

[Maryam483]

Hai,
I have trained supervised model just like steps you give (step 1-4), after training supervised baseline model on COCO dataset,
I have run semi_dest.sh with corresponding file paths to determine performance of supervised model, and the performance is
12% ( I have used 10% as partially labelled data) but in Table 1 of your paper the result is "23.70 ± 0.22". How I solve this issue??

Secondly, I am training model on 1 GPU device. This is the only difference.

I am waiting for your positive response and guidance please. Thanks.

[chenbinghui1]

Hai, I have trained supervised model just like steps you give (step 1-4), after training supervised baseline model on COCO dataset, I have run semi_dest.sh with corresponding file paths to determine performance of supervised model, and the performance is 12% ( I have used 10% as partially labelled data) but in Table 1 of your paper the result is "23.70 ± 0.22". How I solve this issue??

Secondly, I am training model on 1 GPU device. This is the only difference.

I am waiting for your positive response and guidance please. Thanks.

Learning rate and running epochs might be adjusted for different GPU numbers.

[Maryam483]

I did not understand should I kept learning rate low or high, with one GPU device????
Because, I have result of 12% and your results is 23%. I have added code of config file below? Please tell me which parameters should I tune???
Secondly, does GPU devices matter in achieving same performance like you used 8 GPUs, but i have used 1 GPU.??

==================================
Code

**# model settings
model = dict(
type='FCOS',
backbone=dict(
type='ResNet',
depth=50,
num_stages=4,
out_indices=(0, 1, 2, 3),
frozen_stages=1,
norm_cfg=dict(type='BN', requires_grad=False),
norm_eval=True,
style='caffe',
init_cfg=dict(
type='Pretrained',
checkpoint='open-mmlab://detectron2/resnet50_caffe')),
neck=dict(
type='FPN',
in_channels=[256, 512, 1024, 2048],
out_channels=256,
start_level=1,
add_extra_convs='on_output', # use P5
num_outs=5,
relu_before_extra_convs=True),
bbox_head=dict(
type='FCOSHead',
num_classes=80,
in_channels=256,
stacked_convs=4,
feat_channels=256,
strides=[8, 16, 32, 64, 128],
norm_on_bbox=True,
centerness_on_reg=True,
dcn_on_last_conv=False,
center_sampling=True,
conv_bias=True,
loss_cls=dict(
type='FocalLoss',
use_sigmoid=True,
gamma=2.0,
alpha=0.25,
loss_weight=1.0),
loss_bbox=dict(type='GIoULoss', loss_weight=1.0),
loss_centerness=dict(
type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0)),
# training and testing settings
train_cfg=dict(
assigner=dict(
type='MaxIoUAssigner',
pos_iou_thr=0.5,
neg_iou_thr=0.4,
min_pos_iou=0,
ignore_iof_thr=-1),
allowed_border=-1,
pos_weight=-1,
debug=False),
test_cfg=dict(
nms_pre=1000,
min_bbox_size=0,
score_thr=0.05,
nms=dict(type='nms', iou_threshold=0.5),
max_per_img=100)
)

img_norm_cfg = dict(
mean=[103.530, 116.280, 123.675], std=[1.0, 1.0, 1.0], to_rgb=False)
train_pipeline = [
dict(type='LoadImageFromFile'),
dict(type='LoadAnnotations', with_bbox=True),
dict(
type='Resize',
img_scale=[(1333, 640), (1333, 800)],
multiscale_mode='value',
keep_ratio=True),
dict(type='RandomFlip', flip_ratio=0.5),
dict(type='Normalize', **img_norm_cfg),
dict(type='Pad', size_divisor=32),
dict(type='DefaultFormatBundle'),
dict(type='Collect', keys=['img', 'gt_bboxes', 'gt_labels']),
]
test_pipeline = [
dict(type='LoadImageFromFile'),
dict(
type='MultiScaleFlipAug',
img_scale=(1333, 800),
flip=False,
transforms=[
dict(type='Resize', keep_ratio=True),
dict(type='RandomFlip'),
dict(type='Normalize', **img_norm_cfg),
dict(type='Pad', size_divisor=32),
dict(type='ImageToTensor', keys=['img']),
dict(type='Collect', keys=['img']),
])
]

dataset_type = 'CocoDataset'

recommend to use absolute path

data_root = '/gruntdata1/bhchen/factory/data/semicoco/'
data_root="I:/DSL4/data/semicoco/"
data = dict(
samples_per_gpu=2,
workers_per_gpu=2,
train=dict(
type=dataset_type,
ann_file='I:/DSL4/DSL-main/data_list/coco_semi/semi_supervised/[email protected]',
img_prefix=data_root + 'images/full/',
pipeline=train_pipeline),
val=dict(
type=dataset_type,
ann_file='I:/DSL4/DSL-main/data_list/coco_semi/semi_supervised/instances_val2017.json',
img_prefix=data_root + 'valid_images/full/',
pipeline=test_pipeline),
## This bellow is used for generating the initial pseudo-labels for unlabeled images when running tools/inference_unlabeled_coco_data.sh
test=dict(
type=dataset_type,
ann_file='I:/DSL4/DSL-main/data_list/coco_semi/semi_supervised/[email protected]',
img_prefix=data_root + 'images/full/',
pipeline=test_pipeline)
## This bellow is used for testing model performances over COCO dataset, if you use toos/semi_dist_test.sh, please uncomment this bellow codes and comment the above
#test=dict(
# type = dataset_type,
# ann_file = 'data_list/coco_semi/semi_supervised/instances_val2017.json',
# img_prefix = data_root + 'valid_images/full/',
# pipeline = test_pipeline,
# )
)

learning policy

optimizer = dict(type='SGD', lr=0.01, momentum=0.9, weight_decay=0.0001,paramwise_cfg=dict(bias_lr_mult=2., bias_decay_mult=0.))
optimizer_config = dict(
#delete=True,
grad_clip=None)

learning policy

lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=1.0 / 3,
# 0.01/0.02 data 150-180-200; 0.05 data 90-120-140; 0.1data 50-80-100; full data 16-22-24
step=[50, 80])
runner = dict(type='EpochBasedRunner', max_epochs=100)
evaluation = dict(interval=1, metric='bbox')

checkpoint_config = dict(interval=5)

yapf:disable

log_config = dict(
interval=10,
hooks=[
dict(type='TextLoggerHook'),
# dict(type='TensorboardLoggerHook')
])

yapf:enable

custom_hooks = [dict(type='NumClassCheckHook')]

dist_params = dict(backend='nccl')
log_level = 'INFO'
load_from = None
resume_from = None
workflow = [('train', 1)]**

[chenbinghui1]

@Maryam483 Sorry, I didn't test on 1 gpu. So I cannot tell you how to change the value. You can try by your self, good luck~

[Maryam483]

The semi-supervised model requires 158 days on my "1 GPU device" . Why??

You have used 8 GPUs but I have 1 GPU device.

I have "Titan XP (12GB VRAM) GPU with 16 GB RAM". I have added screenshot below.

[chenbinghui1]

@Maryam483 ;
Please check
(1) Whether moving model to GPU for computing

If yes, emm, you can change a task. Titan Xp might not be good for detection task.

I use 8 NVIDIA-V100 for training