为什么使用cbam训练后，显示map有提升，但是推理时出现了预测图片没有效果（整张图片没有预测框）的问题

Question

为什么使用cbam训练后，显示map有提升，但是推理时出现了预测图片没有效果（整张图片没有预测框）的问题

Opened this issue a month ago · 0 comments

Prerequisite

I have searched the existing and past issues but cannot get the expected help.
I have read the FAQ documentation but cannot get the expected help.
The bug has not been fixed in the latest version.

💬 Describe the reimplementation questions

代码修改：
我添加了cbam插件
model = dict(
backbone=dict(
plugins=[
dict(cfg=dict(type='CBAM'),
stages=(True, True, True, True)),
]),
bbox_head=dict(
head_module=dict(num_classes=num_classes),
prior_generator=dict(base_sizes=anchors),

    # loss_cls 会根据 num_classes 动态调整，但是 num_classes = 1 的时候，loss_cls 恒为 0
    loss_cls=dict(loss_weight=0.5 *
                  (num_classes / 80 * 3 / _base_.num_det_layers))))

报告：
!!!You are using YOLOv5Head with num_classes == 1. The loss_cls will be 0. This is a normal phenomenon.
Loads checkpoint by local backend from path: work_dirs_LLVIP/with_cbam/best_coco_bbox_mAP_epoch_44.pth
The model and loaded state dict do not match exactly

unexpected key in source state_dict: backbone.stage1.2.channel_attention.fc.0.conv.weight, backbone.stage1.2.channel_attention.fc.0.conv.bias, backbone.stage1.2.channel_attention.fc.1.conv.weight, backbone.stage1.2.channel_attention
.fc.1.conv.bias, backbone.stage1.2.spatial_attention.conv.conv.weight, backbone.stage1.2.spatial_attention.conv.conv.bias, backbone.stage4.3.channel_attention.fc.0.conv.weight, backbone.stage4.3.channel_attention.fc.0.conv.bias, bac
kbone.stage4.3.channel_attention.fc.1.conv.weight, backbone.stage4.3.channel_attention.fc.1.conv.bias, backbone.stage4.3.spatial_attention.conv.conv.weight, backbone.stage4.3.spatial_attention.conv.conv.bias

F:\Miniconda3\envs\mmyolo2\lib\site-packages\mmengine\visualization\visualizer.py:196: UserWarning: Failed to add <class 'mmengine.visualization.vis_backend.LocalVisBackend'>, please provide the save_dir argument.
warnings.warn(f'Failed to add {vis_backend.class}, '
F:\Miniconda3\envs\mmyolo2\lib\site-packages\mmengine\visualization\visualizer.py:196: UserWarning: Failed to add <class 'mmengine.visualization.vis_backend.TensorboardVisBackend'>, please provide the save_dir argument.
warnings.warn(f'Failed to add {vis_backend.class}, '
[>>>>>>>>>>>>>>>>>>>>>> ] 9/20, 1.8 task/s, elapsed: 5s, ETA: 6sF:\Miniconda3\envs\mmyolo2\lib\site-packages\mmdet\visualization\palette.py:90: UserWarning: floordiv is deprecated, and its behavior
will change in a future version of pytorch. It currently rounds toward 0 (like the 'trunc' function NOT 'floor'). This results in incorrect rounding for negative values. To keep the current behavior, use torch.div(a, b, rounding_mo
de='trunc'), or for actual floor division, use torch.div(a, b, rounding_mode='floor').
scales = 0.5 + (areas - min_area) // (max_area - min_area)
[>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>] 20/20, 2.7 task/s, elapsed: 7s, ETA: 0s
Results have been saved at F:\ssq\data_small_test\out

Environment

absl-py 2.1.0 pypi_0 pypi
addict 2.4.0 pypi_0 pypi
albumentations 1.4.2 pypi_0 pypi
aliyun-python-sdk-core 2.15.0 pypi_0 pypi
aliyun-python-sdk-kms 2.16.2 pypi_0 pypi
appdirs 1.4.4 pypi_0 pypi
bcrypt 4.1.2 pypi_0 pypi
ca-certificates 2024.3.11 haa95532_0 defaults
cachetools 5.3.3 pypi_0 pypi
certifi 2024.2.2 pypi_0 pypi
cffi 1.16.0 pypi_0 pypi
charset-normalizer 3.3.2 pypi_0 pypi
cityscapesscripts 2.2.2 pypi_0 pypi
click 8.1.7 pypi_0 pypi
cloudpickle 3.0.0 pypi_0 pypi
colorama 0.4.6 pypi_0 pypi
coloredlogs 15.0.1 pypi_0 pypi
contourpy 1.1.1 pypi_0 pypi
crcmod 1.7 pypi_0 pypi
cryptography 42.0.5 pypi_0 pypi
cycler 0.12.1 pypi_0 pypi
filelock 3.9.0 pypi_0 pypi
fonttools 4.50.0 pypi_0 pypi
fsspec 2023.4.0 pypi_0 pypi
google-auth 2.28.2 pypi_0 pypi
google-auth-oauthlib 1.0.0 pypi_0 pypi
grpcio 1.62.1 pypi_0 pypi
gym 0.26.2 pypi_0 pypi
gym-notices 0.0.8 pypi_0 pypi
humanfriendly 10.0 pypi_0 pypi
idna 3.6 pypi_0 pypi
imageio 2.34.0 pypi_0 pypi
importlib-metadata 7.0.2 pypi_0 pypi
importlib-resources 6.3.1 pypi_0 pypi
jinja2 3.1.2 pypi_0 pypi
jmespath 0.10.0 pypi_0 pypi
joblib 1.3.2 pypi_0 pypi
kiwisolver 1.4.5 pypi_0 pypi
lazy-loader 0.3 pypi_0 pypi
libffi 3.4.4 hd77b12b_0 defaults
markdown 3.6 pypi_0 pypi
markdown-it-py 3.0.0 pypi_0 pypi
markupsafe 2.1.5 pypi_0 pypi
matplotlib 3.7.5 pypi_0 pypi
mdurl 0.1.2 pypi_0 pypi
mmcv 2.0.1 pypi_0 pypi
mmdet 3.3.0 pypi_0 pypi
mmengine 0.10.3 pypi_0 pypi
mmyolo 0.6.0 pypi_0 pypi
model-index 0.1.11 pypi_0 pypi
mpmath 1.3.0 pypi_0 pypi
nes-py 8.2.1 pypi_0 pypi
networkx 3.1 pypi_0 pypi
numpy 1.24.4 pypi_0 pypi
oauthlib 3.2.2 pypi_0 pypi
opencv-python 4.9.0.80 pypi_0 pypi
opencv-python-headless 4.9.0.80 pypi_0 pypi
opendatalab 0.0.10 pypi_0 pypi
openmim 0.3.9 pypi_0 pypi
openssl 3.0.13 h2bbff1b_0 defaults
openxlab 0.0.36 pypi_0 pypi
ordered-set 4.1.0 pypi_0 pypi
oss2 2.17.0 pypi_0 pypi
packaging 24.0 pypi_0 pypi
pandas 2.0.3 pypi_0 pypi
pillow 10.2.0 pypi_0 pypi
pip 24.0 pypi_0 pypi
platformdirs 4.2.0 pypi_0 pypi
prettytable 3.10.0 pypi_0 pypi
protobuf 5.26.0 pypi_0 pypi
pyasn1 0.5.1 pypi_0 pypi
pyasn1-modules 0.3.0 pypi_0 pypi
pycocotools 2.0.7 pypi_0 pypi
pycparser 2.21 pypi_0 pypi
pycryptodome 3.20.0 pypi_0 pypi
pyglet 1.5.21 pypi_0 pypi
pygments 2.17.2 pypi_0 pypi
pynacl 1.5.0 pypi_0 pypi
pyparsing 3.1.2 pypi_0 pypi
pyquaternion 0.9.9 pypi_0 pypi
pyreadline3 3.4.1 pypi_0 pypi
python 3.8.18 h1aa4202_0 defaults
python-dateutil 2.9.0.post0 pypi_0 pypi
pytz 2023.4 pypi_0 pypi
pywavelets 1.4.1 pypi_0 pypi
pywin32 306 pypi_0 pypi
pyyaml 6.0.1 pypi_0 pypi
regex 2023.12.25 pypi_0 pypi
requests 2.28.2 pypi_0 pypi
requests-oauthlib 1.4.0 pypi_0 pypi
rich 13.4.2 pypi_0 pypi
rsa 4.9 pypi_0 pypi
scikit-image 0.21.0 pypi_0 pypi
scikit-learn 1.3.2 pypi_0 pypi
scipy 1.10.1 pypi_0 pypi
setuptools 60.2.0 pypi_0 pypi
shapely 2.0.3 pypi_0 pypi
six 1.16.0 pypi_0 pypi
sqlite 3.41.2 h2bbff1b_0 defaults
sympy 1.12 pypi_0 pypi
tabulate 0.9.0 pypi_0 pypi
tensorboard 2.14.0 pypi_0 pypi
tensorboard-data-server 0.7.2 pypi_0 pypi
termcolor 2.4.0 pypi_0 pypi
terminaltables 3.1.10 pypi_0 pypi
threadpoolctl 3.3.0 pypi_0 pypi
tifffile 2023.7.10 pypi_0 pypi
tomli 2.0.1 pypi_0 pypi
torch 1.12.1+cu113 pypi_0 pypi
torchaudio 0.12.1+cu113 pypi_0 pypi
torchvision 0.13.1+cu113 pypi_0 pypi
tqdm 4.65.2 pypi_0 pypi
typing 3.7.4.3 pypi_0 pypi
typing-extensions 4.10.0 pypi_0 pypi
tzdata 2024.1 pypi_0 pypi
urllib3 1.26.18 pypi_0 pypi
vc 14.2 h21ff451_1 defaults
vs2015_runtime 14.27.29016 h5e58377_2 defaults
wcwidth 0.2.13 pypi_0 pypi
werkzeug 3.0.1 pypi_0 pypi
wheel 0.43.0 pypi_0 pypi
yapf 0.40.2 pypi_0 pypi
zipp 3.18.1 pypi_0 pypi

Expected results

我想知道为什么推理的结果是识别不出我的类，并且我想知道cbam应该在哪个阶段使用，我在3,4阶段使用，1,2阶段不使用时效果反而出现了下降

Additional information

我仅仅添加了插件，未添加前的推理结果是可观的