Question about SSD on VOC

[ghost]

Hello,

Thank you for sharing works!

I ran MIAOD_SSD.py config to reproduce the Figure 5(b) in the paper.
But, the number of labeled images increases by 2k (not 1k) each cycle.

[yuantn]

Hello,

Thanks for your attention to our work!

There seems to be no problems in the config. Could you provide your output log for examination?

[ghost]

2021-09-06 09:44:26,951 - mmdet - INFO - Environment info:

sys.platform: linux
Python: 3.6.12 (default, Aug 18 2020, 02:08:22) [GCC 5.4.0 20160609]
CUDA available: True
CUDA_HOME: /usr/local/cuda
NVCC: Cuda compilation tools, release 10.1, V10.1.243
GPU 0: Tesla P40
GCC: gcc (Ubuntu 5.4.0-6ubuntu1~16.04.12) 5.4.0 20160609
PyTorch: 1.6.0
PyTorch compiling details: PyTorch built with:

• GCC 7.3
• C++ Version: 201402
• Intel(R) Math Kernel Library Version 2019.0.5 Product Build 20190808 for Intel(R) 64 architecture applications
• Intel(R) MKL-DNN v1.5.0 (Git Hash e2ac1fac44c5078ca927cb9b90e1b3066a0b2ed0)
• OpenMP 201511 (a.k.a. OpenMP 4.5)
• NNPACK is enabled
• CPU capability usage: AVX2
• CUDA Runtime 10.2
• NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
• CuDNN 7.6.5
• Magma 2.5.2
• Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Wer$

TorchVision: 0.7.0
OpenCV: 4.5.3
MMCV: 1.0.5
MMDetection: 2.3.0+e343491
MMDetection Compiler: GCC 5.4
MMDetection CUDA Compiler: 10.1

2021-09-06 09:44:26,952 - mmdet - INFO - Distributed training: False
2021-09-06 09:44:27,630 - mmdet - INFO - Config:
input_size = 300
model = dict(
type='SingleStageDetector',
pretrained='open-mmlab://vgg16_caffe',
backbone=dict(
type='SSDVGG',
input_size=300,
depth=16,
with_last_pool=False,
ceil_mode=True,
out_indices=(3, 4),
out_feature_indices=(22, 34),
l2_norm_scale=20),
neck=None,
bbox_head=dict(
type='SSDHead',
in_channels=(512, 1024, 512, 256, 256, 256),
C=20,
anchor_generator=dict(
type='SSDAnchorGenerator',
scale_major=False,
input_size=300,
basesize_ratio_range=(0.2, 0.9),
strides=[8, 16, 32, 64, 100, 300],
ratios=[[2], [2, 3], [2, 3], [2, 3], [2], [2]]),
bbox_coder=dict(
type='DeltaXYWHBBoxCoder',
target_means=[0.0, 0.0, 0.0, 0.0],
target_stds=[0.1, 0.1, 0.2, 0.2])))
cudnn_benchmark = True
train_cfg = dict(
assigner=dict(
type='MaxIoUAssigner',
pos_iou_thr=0.5,
neg_iou_thr=0.5,
min_pos_iou=0.0,
ignore_iof_thr=-1,
gt_max_assign_all=False),
smoothl1_beta=1.0,
allowed_border=-1,
pos_weight=-1,
neg_pos_ratio=3,
debug=False,
param_lambda=0.5)
test_cfg = dict(
nms=dict(type='nms', iou_threshold=0.45),
min_bbox_size=0,
score_thr=0.02,
max_per_img=200)
theta_f_1 = [
'bbox_head.f_1_convs.0.weight', 'bbox_head.f_1_convs.0.bias',
'bbox_head.f_1_convs.1.weight', 'bbox_head.f_1_convs.1.bias',
'bbox_head.f_1_convs.2.weight', 'bbox_head.f_1_convs.2.bias',
'bbox_head.f_1_convs.3.weight', 'bbox_head.f_1_convs.3.bias',
'bbox_head.f_1_convs.4.weight', 'bbox_head.f_1_convs.4.bias',
'bbox_head.f_1_convs.5.weight', 'bbox_head.f_1_convs.5.bias'
]
theta_f_2 = [
'bbox_head.f_2_convs.0.weight', 'bbox_head.f_2_convs.0.bias',
'bbox_head.f_2_convs.1.weight', 'bbox_head.f_2_convs.1.bias',
'bbox_head.f_2_convs.2.weight', 'bbox_head.f_2_convs.2.bias',
'bbox_head.f_2_convs.3.weight', 'bbox_head.f_2_convs.3.bias',
'bbox_head.f_2_convs.4.weight', 'bbox_head.f_2_convs.4.bias',
'bbox_head.f_2_convs.5.weight', 'bbox_head.f_2_convs.5.bias'
]
data_root = '/muzi/data/al/od/VOCdevkit/'
dataset_type = 'VOCDataset'
img_norm_cfg = dict(mean=[123.675, 116.28, 103.53], std=[1, 1, 1], to_rgb=True)
train_pipeline = [
dict(type='LoadImageFromFile', to_float32=True),
dict(type='LoadAnnotations', with_bbox=True),
dict(
type='PhotoMetricDistortion',
brightness_delta=32,
contrast_range=(0.5, 1.5),
saturation_range=(0.5, 1.5),
hue_delta=18),
dict(
type='Expand',
mean=[123.675, 116.28, 103.53],
to_rgb=True,
ratio_range=(1, 4)),
dict(
type='MinIoURandomCrop',
min_ious=(0.1, 0.3, 0.5, 0.7, 0.9),
min_crop_size=0.3),
dict(type='Resize', img_scale=(300, 300), keep_ratio=False),
dict(
type='Normalize',
mean=[123.675, 116.28, 103.53],
std=[1, 1, 1],
to_rgb=True),
dict(type='RandomFlip', flip_ratio=0.5),
dict(type='DefaultFormatBundle'),
dict(type='Collect', keys=['img', 'gt_bboxes', 'gt_labels'])
]
test_pipeline = [
dict(type='LoadImageFromFile'),
dict(
type='MultiScaleFlipAug',
img_scale=(300, 300),
flip=False,
transforms=[
dict(type='Resize', keep_ratio=False),
dict(
type='Normalize',
mean=[123.675, 116.28, 103.53],
std=[1, 1, 1],
to_rgb=True),
dict(type='ImageToTensor', keys=['img']),
dict(type='Collect', keys=['img'])
])
]
data = dict(
samples_per_gpu=8,
workers_per_gpu=3,
train=dict(
type='RepeatDataset',
times=1,
dataset=dict(
type='VOCDataset',
ann_file=[
'/muzi/data/al/od/VOCdevkit/VOC2007/ImageSets/Main/trainval.txt',
'/muzi/data/al/od/VOCdevkit/VOC2012/ImageSets/Main/trainval.txt'
],
img_prefix=[
'/muzi/data/al/od/VOCdevkit/VOC2007/',
'/muzi/data/al/od/VOCdevkit/VOC2012/'
],
pipeline=[
dict(type='LoadImageFromFile', to_float32=True),
dict(type='LoadAnnotations', with_bbox=True),
dict(
type='PhotoMetricDistortion',
brightness_delta=32,
contrast_range=(0.5, 1.5),
saturation_range=(0.5, 1.5),
hue_delta=18),
dict(
type='Expand',
mean=[123.675, 116.28, 103.53],
to_rgb=True,
ratio_range=(1, 4)),
dict(
type='MinIoURandomCrop',
min_ious=(0.1, 0.3, 0.5, 0.7, 0.9),
min_crop_size=0.3),
dict(type='Resize', img_scale=(300, 300), keep_ratio=False),
dict(
type='Normalize',
mean=[123.675, 116.28, 103.53],
std=[1, 1, 1],
to_rgb=True),
dict(type='RandomFlip', flip_ratio=0.5),
dict(type='DefaultFormatBundle'),
dict(type='Collect', keys=['img', 'gt_bboxes', 'gt_labels'])
])),
val=dict(
type='VOCDataset',
ann_file='/muzi/data/al/od/VOCdevkit/VOC2007/ImageSets/Main/test.txt',
img_prefix='/muzi/data/al/od/VOCdevkit/VOC2007/',
pipeline=[
dict(type='LoadImageFromFile'),
dict(
type='MultiScaleFlipAug',
img_scale=(300, 300),
flip=False,
transforms=[
dict(type='Resize', keep_ratio=False),
dict(
type='Normalize',
mean=[123.675, 116.28, 103.53],
std=[1, 1, 1],
to_rgb=True),
dict(type='ImageToTensor', keys=['img']),
dict(type='Collect', keys=['img'])
])
]),
test=dict(
type='VOCDataset',
ann_file=[
'/muzi/data/al/od/VOCdevkit/VOC2007/ImageSets/Main/trainval.txt',
'/muzi/data/al/od/VOCdevkit/VOC2012/ImageSets/Main/trainval.txt'
],
img_prefix=[
'/muzi/data/al/od/VOCdevkit/VOC2007/',
'/muzi/data/al/od/VOCdevkit/VOC2012/'
],
pipeline=[
dict(type='LoadImageFromFile'),
dict(
type='MultiScaleFlipAug',
img_scale=(300, 300),
flip=False,
transforms=[
dict(type='Resize', keep_ratio=False),
dict(
type='Normalize',
mean=[123.675, 116.28, 103.53],
std=[1, 1, 1],
to_rgb=True),
dict(type='ImageToTensor', keys=['img']),
dict(type='Collect', keys=['img'])
])
]))
evaluation = dict(interval=5, metric='mAP')
checkpoint_config = dict(interval=1)
log_config = dict(interval=50, hooks=[dict(type='TextLoggerHook')])
dist_params = dict(backend='nccl')
log_level = 'INFO'
load_from = None
resume_from = None
workflow = [('train', 1)]
optimizer = dict(type='SGD', lr=0.001, momentum=0.9, weight_decay=0.0005)
optimizer_config = dict()
lr_config = dict(
policy='step',
warmup='linear',
warmup_iters=500,
warmup_ratio=0.001,
step=[1])
epoch_ratio = [5, 1]
epoch = 2
X_L_repeat = 16
X_U_repeat = 16
k = 10000
X_S_size = 1000
X_L_0_size = 1000
cycles = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
work_directory = '/muzi/logs/al/od/miaod/voc_ssd/trial_0'
gpu_ids = range(0, 1)

2021-09-06 09:44:27,631 - mmdet - INFO - Set random seed to 666, deterministic: False
2021-09-06 09:44:27,715 - mmdet - INFO - Set random seed to 666, deterministic: False
2021-09-06 09:44:28,085 - mmdet - INFO - load model from: open-mmlab://vgg16_caffe
2021-09-06 09:44:28,161 - mmdet - WARNING - The model and loaded state dict do not match exactly

missing keys in source state_dict: extra.0.weight, extra.0.bias, extra.1.weight, extra.1.bias, extra.2.weight, extra.2.bias, extra.3.weight, extra.3.bias, extra.4.weight, extra.4.bias, extra.5.weight, extra.5.bias, extra.6.weight, extra.6.bias, extra.7.weight, extra.7$

2021-09-06 09:44:37,441 - mmdet - INFO - Start running, host: svcapp@workspace-899, work_directory: /muzi/logs/al/od/miaod/voc_ssd/trial_0/20210906_094426
2021-09-06 09:44:37,442 - mmdet - INFO - workflow: [('train', 1)], max: 5 epochs
2021-09-06 09:45:01,958 - mmdet - INFO - Epoch [1][50/2000] lr: 9.890e-05, eta: 1:20:52, time: 0.488, data_time: 0.056, memory: 10018, l_det_cls: 17.9104, l_det_loc: 3.0110, l_imgcls: 0.2427, L_det: 21.1641
2021-09-06 09:45:20,605 - mmdet - INFO - Epoch [1][100/2000] lr: 1.988e-04, eta: 1:10:59, time: 0.373, data_time: 0.012, memory: 10018, l_det_cls: 12.3032, l_det_loc: 2.7717, l_imgcls: 0.2481, L_det: 15.3230
2021-09-06 09:45:39,274 - mmdet - INFO - Epoch [1][150/2000] lr: 2.987e-04, eta: 1:07:31, time: 0.373, data_time: 0.012, memory: 10018, l_det_cls: 10.8066, l_det_loc: 2.8618, l_imgcls: 0.2456, L_det: 13.9141
2021-09-06 09:45:58,018 - mmdet - INFO - Epoch [1][200/2000] lr: 3.986e-04, eta: 1:05:41, time: 0.375, data_time: 0.012, memory: 10018, l_det_cls: 6.7048, l_det_loc: 2.7885, l_imgcls: 0.2420, L_det: 9.7354
2021-09-06 09:46:16,628 - mmdet - INFO - Epoch [1][250/2000] lr: 4.985e-04, eta: 1:04:22, time: 0.372, data_time: 0.012, memory: 10018, l_det_cls: 5.6036, l_det_loc: 2.6494, l_imgcls: 0.2207, L_det: 8.4737
2021-09-06 09:46:35,303 - mmdet - INFO - Epoch [1][300/2000] lr: 5.984e-04, eta: 1:03:26, time: 0.373, data_time: 0.012, memory: 10018, l_det_cls: 5.1834, l_det_loc: 2.6133, l_imgcls: 0.2228, L_det: 8.0195
2021-09-06 09:46:54,704 - mmdet - INFO - Epoch [1][350/2000] lr: 6.983e-04, eta: 1:03:00, time: 0.388, data_time: 0.012, memory: 10018, l_det_cls: 5.3418, l_det_loc: 2.6069, l_imgcls: 0.2182, L_det: 8.1669
2021-09-06 09:47:13,962 - mmdet - INFO - Epoch [1][400/2000] lr: 7.982e-04, eta: 1:02:33, time: 0.385, data_time: 0.012, memory: 10018, l_det_cls: 5.2064, l_det_loc: 2.4717, l_imgcls: 0.2175, L_det: 7.8956
2021-09-06 09:47:32,623 - mmdet - INFO - Epoch [1][450/2000] lr: 8.981e-04, eta: 1:01:54, time: 0.373, data_time: 0.012, memory: 10018, l_det_cls: 5.0469, l_det_loc: 2.4876, l_imgcls: 0.2173, L_det: 7.7519
2021-09-06 09:47:51,335 - mmdet - INFO - Epoch [1][500/2000] lr: 9.980e-04, eta: 1:01:21, time: 0.374, data_time: 0.012, memory: 10018, l_det_cls: 5.0143, l_det_loc: 2.3822, l_imgcls: 0.2201, L_det: 7.6166
2021-09-06 09:48:10,087 - mmdet - INFO - Epoch [1][550/2000] lr: 1.000e-03, eta: 1:00:51, time: 0.375, data_time: 0.012, memory: 10018, l_det_cls: 5.1041, l_det_loc: 2.3307, l_imgcls: 0.2302, L_det: 7.6649
2021-09-06 09:48:28,805 - mmdet - INFO - Epoch [1][600/2000] lr: 1.000e-03, eta: 1:00:22, time: 0.374, data_time: 0.012, memory: 10018, l_det_cls: 5.1053, l_det_loc: 2.2741, l_imgcls: 0.2200, L_det: 7.5993
2021-09-06 09:48:47,365 - mmdet - INFO - Epoch [1][650/2000] lr: 1.000e-03, eta: 0:59:53, time: 0.371, data_time: 0.012, memory: 10018, l_det_cls: 5.0543, l_det_loc: 2.2532, l_imgcls: 0.2153, L_det: 7.5227
2021-09-06 09:49:05,948 - mmdet - INFO - Epoch [1][700/2000] lr: 1.000e-03, eta: 0:59:25, time: 0.372, data_time: 0.012, memory: 10018, l_det_cls: 5.0043, l_det_loc: 2.2249, l_imgcls: 0.2247, L_det: 7.4539
2021-09-06 09:49:24,792 - mmdet - INFO - Epoch [1][750/2000] lr: 1.000e-03, eta: 0:59:02, time: 0.377, data_time: 0.012, memory: 10018, l_det_cls: 4.8676, l_det_loc: 2.1896, l_imgcls: 0.2192, L_det: 7.2765
2021-09-06 09:49:43,671 - mmdet - INFO - Epoch [1][800/2000] lr: 1.000e-03, eta: 0:58:40, time: 0.378, data_time: 0.012, memory: 10018, l_det_cls: 4.9734, l_det_loc: 2.1614, l_imgcls: 0.2261, L_det: 7.3610
2021-09-06 09:50:02,444 - mmdet - INFO - Epoch [1][850/2000] lr: 1.000e-03, eta: 0:58:17, time: 0.375, data_time: 0.013, memory: 10018, l_det_cls: 4.9286, l_det_loc: 2.1764, l_imgcls: 0.2267, L_det: 7.3317
2021-09-06 09:50:21,276 - mmdet - INFO - Epoch [1][900/2000] lr: 1.000e-03, eta: 0:57:55, time: 0.377, data_time: 0.012, memory: 10018, l_det_cls: 4.9132, l_det_loc: 2.1281, l_imgcls: 0.2172, L_det: 7.2585
2021-09-06 09:50:39,892 - mmdet - INFO - Epoch [1][950/2000] lr: 1.000e-03, eta: 0:57:31, time: 0.372, data_time: 0.012, memory: 10018, l_det_cls: 4.9181, l_det_loc: 2.2532, l_imgcls: 0.2215, L_det: 7.3928
2021-09-06 09:50:58,791 - mmdet - INFO - Epoch [1][1000/2000] lr: 1.000e-03, eta: 0:57:10, time: 0.378, data_time: 0.012, memory: 10018, l_det_cls: 4.8111, l_det_loc: 2.1222, l_imgcls: 0.2216, L_det: 7.1549

[yuantn]

Thanks, it seems normal.

And how do you know that the number of labeled images increases by 2k but not 1k each cycle?

Could you provide the shape of X_L_0.npy and X_L_1.npy in /muzi/logs/al/od/miaod/voc_ssd/trial_0/$TIMESTAMP/ (where the $TIMESTAMP is like 20210907_203401) ?

[ghost]

Sorry, I misunderstood that maximum iterations is the number of labeled images in training logs, e.g.
2021-09-06 09:45:01,958 - mmdet - INFO - Epoch [1][50/2000] lr: 9.890e-05, eta: 1:20:52, time: 0.488, data_time: 0.056, memory: 10018, l_det_cls: 17.9104, l_det_loc: 3.0110, l_imgcls: 0.2427, L_det: 21.1641
2000 is calculated by X_L_0_size * X_L_repeat / samples_per_gpu = 1000 * 16 / 8 according to MIAOD_SSD.py, right?

I checked the shape of X_L_0.npy and X_L_1.npy is (1000,) and (2000,).
Thank you very much!

[yuantn]

Yes, that's it.

[Liang-ZX]

Sorry, I misunderstood that maximum iterations is the number of labeled images in training logs, e.g. 2021-09-06 09:45:01,958 - mmdet - INFO - Epoch [1][50/2000] lr: 9.890e-05, eta: 1:20:52, time: 0.488, data_time: 0.056, memory: 10018, l_det_cls: 17.9104, l_det_loc: 3.0110, l_imgcls: 0.2427, L_det: 21.1641 2000 is calculated by X_L_0_size * X_L_repeat / samples_per_gpu = 1000 * 16 / 8 according to MIAOD_SSD.py, right?

I checked the shape of X_L_0.npy and X_L_1.npy is (1000,) and (2000,). Thank you very much!

Excuse me. I would like to know whether your running result of SSD comparable to that in the paper. If so, how do you modify the code? Only the config file? Thanks.

[jiahao12121]

Sorry, I misunderstood that maximum iterations is the number of labeled images in training logs, e.g. 2021-09-06 09:45:01,958 - mmdet - INFO - Epoch [1][50/2000] lr: 9.890e-05, eta: 1:20:52, time: 0.488, data_time: 0.056, memory: 10018, l_det_cls: 17.9104, l_det_loc: 3.0110, l_imgcls: 0.2427, L_det: 21.1641 2000 is calculated by X_L_0_size * X_L_repeat / samples_per_gpu = 1000 * 16 / 8 according to MIAOD_SSD.py, right?
I checked the shape of X_L_0.npy and X_L_1.npy is (1000,) and (2000,). Thank you very much!

Excuse me. I would like to know whether your running result of SSD comparable to that in the paper. If so, how do you modify the code? Only the config file? Thanks.

i meet the same question, and the parameters of my log file is same to the author'. Would you find the reason?