It is a nice work. Will you release the code for reference?

Dear authors,

thanks for this great job.
I'm trying to train the model again, but it requires swin_tiny_patch4_window7_224_d2.pth.
Where do I find these weights?

Thanks for sharing your great work!But I encountered this error for a long time and can't solve it.
Traceback (most recent call last): File "train_net.py", line 33, in <module> from dyhead import add_dyhead_config File "/user-data/DynamicHead-master/dyhead/__init__.py", line 2, in <module> from .dyhead import DyHead File "/user-data/DynamicHead-master/dyhead/dyhead.py", line 7, in <module> from .deform import ModulatedDeformConv File "/user-data/DynamicHead-master/dyhead/deform.py", line 8, in <module> import dyhead._C as _C ModuleNotFoundError: No module named 'dyhead._C'

Could you please give out a dockerfile to set up the environment?
I found setting up the environment is difficult. Thanks.

Hello, I'm very interested in your improvements in the Head section. When will the code be open source?

Thanks!

I have tried 2 different configs with diferent backbones, but all occurs exception of one thing, that is weight_decay is NoneType in optimizer SGD and ADAMW. It seems the configs of WEIGHT_DECAY is not imported correctly?

How about the speed of model(with Swin-L) which achieve 60.6mAP? such as on TITAN X?

hi, thanks for your repo.

Can I in place the deformconv with regular conv?
what's the influence?

thanks!

Hi,
First, thank you for your work!
I tried to train a network on my side and I feel like some modulated convolutions are not working as intended regarding the following code:
temp_fea.append(F.upsample_bilinear(self.DyConv[0](x[feature_names[level + 1]], **conv_args), size=[feature.size(2), feature.size(3)])) (l.71-72 of dyhead.py)
When running this line, the modulated conv receives an input which is four times smaller than the offset and mask (twice shorter on H and W dimensions).
As there is no "assert" on the shape of the inputs, the code runs fine but what is being computed is not really what you expect: the offset and the mask are flattened and only the first quarter of the vector is being used.
This leads to a huge shifting in the computation of the output of the modulated convolution.
To "fix" the issue, I think that the upsample_bilinear() should be applied on x[featurenames[level + 1]] and not the output of the layer.
Hope it helps.

Excuse me，could you tell how to apply this module to YOLOv4? If so, do we need to modify the loss function? Thanks for your answer very much!

How can I train Dynamichead without using the backbone network, As the training process of the table below

Thank you for sharing the code! I'm seeking for details about scale-aware, spatial-aware and task-aware blocks in the source code. However, as i can see in the source code "dyhead/dyhead.py", there only exists the implementation of spatial-aware block.
Is this a mistake or my misunderstanding?

when will we release the model(with Swin-L) which achieve 60.6mAP?

Hello, when I run the train_net.py, I encountered this error:
ImportError: XX/Code/DynamicHead/dyhead/_C.cpython-38-x86_64-linux-gnu.so: undefined symbol: _ZN6caffe28TypeMeta21_typeMetaDataInstanceIdEEPKNS_6detail12TypeMetaDataEv,

It caused by "import dyhead._C as _C", I tried change the torch and torchvision version, but it not works.
May you find the solution?

I want to know how to change my dataset. Can you tell me where to modify it? I've changed the dataset to coco format

Thanks for sharing your great work! But I got a "No module issue" when i run your Code. It occurs in deform.py ,
line 8:
"import dyhead._C as _C".

In figure 2(a) of the paper, the is after , but in the code, it seems that the space scale-aware attention is applied after the deformable conv. Is this correct?

Hi, Sir,
At equation 5, it seems we only have a set of (alpha_1, beta_1, alpha_2, beta_2). Dose it tell us all channel features slide work with same activation?

Is it possible to set (alpha_1_c1, beta_1_c1, alpha_2_c1, beta_2_c1) for channel 1, (alpha_1_c2, beta_1_c2, alpha_2_c2, beta_2_c2) for channel 2... refer to dynamic relu fig. 2-b?

looking forward to having your answer soon.

Thanks,
Patrick

In Equation 4 in your paper, you specify an equation for the spatial-aware attention module which averages across the levels of the rescaled spatial pyramid. It appears this equation would map an element of R^(LxSxC) to R^(SxC). Is this intended? How do you apply multiple DyHead modules after this stage, if the L axis is collapsed?

Hi, thanks for your wonderful work! I have a question that In the ATSS, It uses three scales for prediction, If I want to apply Dynamic head to ATSS, the final prediction scale is one?

I think you want use F.adaptive_max_pool2d() instead of F.interpolate().

Hi.
Thanks for your interesting idea.

I'm using swin-tiny for object detection.
But I wanna use Swin-Large.

I'm sorry but can I get Swin-Large config plz?

Thank you.

I do not find the code for the following description

Given a concatenation of features Fin = fFigL i=1 from
L different levels in a feature pyramid, we can resize consecutive
level features towards the scale of the median level
feature using up-sampling or down-sampling. The re-scaled
feature pyramid can be denoted as a 4-dimensional tensor
F 2 RL�H�W�C, where L represents the number of levels
in the pyramid, H, W, and C represent height, width,
and the number of channels of the median level feature respectively.

as the title.

The paper has been accepted by CVPR2021 for a while. When can code be available?

Hi, when I execute this sentence, an error occurs with DynamicHead.
ERROR: DynamicHead is not a valid editable requirement. It should either be a path to a local p
roject or a VCS URL (beginning with bzr+http, bzr+https, bzr+ssh, bzr+sftp, bzr+ftp, bzr+lp, bz
r+file, git+http, git+https, git+ssh, git+git, git+file, hg+file, hg+http,
hg+https, hg+ssh, hg+static-http, svn+ssh, svn+http, svn+https, svn+svn, svn+file).

1、nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2020 NVIDIA Corporation
Built on Tue_Sep_15_19:10:02_PDT_2020
Cuda compilation tools, release 11.1, V11.1.74
Build cuda_11.1.TC455_06.29069683_0

2、pip list
Package Version Location

detectron2 0.6+cu111
torch 1.10.1+cu111
torchaudio 0.10.1+rocm4.1
torchvision 0.11.2+cu111
tqdm 4.63.1

3、log：
sda/okay/ai/project/ocrquesseg-svr/detectron2/detectron2/structures/boxes.py:158: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:201.)
tensor = torch.as_tensor(tensor, dtype=torch.float32, device=device)
error in modulated_deformable_im2col_cuda: no kernel image is available for execution on the device
error in modulated_deformable_im2col_cuda: no kernel image is available for execution on the device
error in modulated_deformable_im2col_cuda: no kernel image is available for execution on the device
error in modulated_deformable_im2col_cuda: no kernel image is available for execution on the device
/sda/program/anaconda3/envs/detectron2/lib/python3.7/site-packages/torch/nn/functional.py:3847: UserWarning: nn.functional.upsample_bilinear is deprecated. Use nn.functional.interpolate instead.
warnings.warn("nn.functional.upsample_bilinear is deprecated. Use nn.functional.interpolate instead.")
error in modulated_deformable_im2col_cuda: no kernel image is available for execution on the device
error in modulated_deformable_im2col_cuda: no kernel image is available for execution on the device
error in modulated_deformable_im2col_cuda: no kernel image is available for execution on the device
error in modulated_deformable_im2col_cuda: no kernel image is available for execution on the device

error in modulated_deformable_im2col_cuda: no kernel image is available for execution on the device
error in modulated_deformable_im2col_cuda: no kernel image is available for execution on the device
error in modulated_deformable_im2col_cuda: no kernel image is available for execution on the device
THCudaCheck FAIL file=/sda/okay/ai/project/ocrquesseg-svr/dyhead/csrc/cuda/SigmoidFocalLoss_cuda.cu line=139 error=209 : no kernel image is available for execution on the device
ERROR [04/01 18:08:11 d2.engine.train_loop]: Exception during training:
Traceback (most recent call last):
File "/sda/okay/ai/project/ocrquesseg-svr/detectron2/detectron2/engine/train_loop.py", line 149, in train
self.run_step()
File "/sda/okay/ai/project/ocrquesseg-svr/detectron2/detectron2/engine/defaults.py", line 499, in run_step
self._trainer.run_step()
File "/sda/okay/ai/project/ocrquesseg-svr/detectron2/detectron2/engine/train_loop.py", line 273, in run_step
loss_dict = self.model(data)
File "/sda/program/anaconda3/envs/detectron2/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/sda/okay/ai/project/ocrquesseg-svr/extra/atss.py", line 204, in forward
losses = self.losses(anchors, pred_logits, gt_labels, pred_anchor_deltas, gt_boxes, pred_centers)
File "/sda/okay/ai/project/ocrquesseg-svr/extra/atss.py", line 236, in losses
cls_loss = self.classification_loss_func(box_cls_flatten, labels_flatten.int()) / num_pos_avg_per_gpu
File "/sda/program/anaconda3/envs/detectron2/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/sda/okay/ai/project/ocrquesseg-svr/extra/sigmoid_focal_loss.py", line 48, in forward
loss = sigmoid_focal_loss_cuda(logits, targets, self.gamma, self.alpha)
File "/sda/okay/ai/project/ocrquesseg-svr/extra/sigmoid_focal_loss.py", line 20, in forward
logits, targets, num_classes, gamma, alpha
RuntimeError: cuda runtime error (209) : no kernel image is available for execution on the device at /sda/okay/ai/project/ocrquesseg-svr/dyhead/csrc/cuda/SigmoidFocalLoss_cuda.cu:139
[04/01 18:08:11 d2.engine.hooks]: Total training time: 0:00:00 (0:00:00 on hooks)
[04/01 18:08:11 d2.utils.events]: iter: 0 lr: N/A max_mem: 6574M
Traceback (most recent call last):
File "train_net.py", line 224, in
args=(args,),
File "/sda/okay/ai/project/ocrquesseg-svr/detectron2/detectron2/engine/launch.py", line 82, in launch
main_func(*args)
File "train_net.py", line 210, in main
return trainer.train()
File "/sda/okay/ai/project/ocrquesseg-svr/detectron2/detectron2/engine/defaults.py", line 489, in train
super().train(self.start_iter, self.max_iter)
File "/sda/okay/ai/project/ocrquesseg-svr/detectron2/detectron2/engine/train_loop.py", line 149, in train
self.run_step()
File "/sda/okay/ai/project/ocrquesseg-svr/detectron2/detectron2/engine/defaults.py", line 499, in run_step
self._trainer.run_step()
File "/sda/okay/ai/project/ocrquesseg-svr/detectron2/detectron2/engine/train_loop.py", line 273, in run_step
loss_dict = self.model(data)
File "/sda/program/anaconda3/envs/detectron2/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/sda/okay/ai/project/ocrquesseg-svr/extra/atss.py", line 204, in forward
losses = self.losses(anchors, pred_logits, gt_labels, pred_anchor_deltas, gt_boxes, pred_centers)
File "/sda/okay/ai/project/ocrquesseg-svr/extra/atss.py", line 236, in losses
cls_loss = self.classification_loss_func(box_cls_flatten, labels_flatten.int()) / num_pos_avg_per_gpu
File "/sda/program/anaconda3/envs/detectron2/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/sda/okay/ai/project/ocrquesseg-svr/extra/sigmoid_focal_loss.py", line 48, in forward
loss = sigmoid_focal_loss_cuda(logits, targets, self.gamma, self.alpha)
File "/sda/okay/ai/project/ocrquesseg-svr/extra/sigmoid_focal_loss.py", line 20, in forward
logits, targets, num_classes, gamma, alpha
RuntimeError: cuda runtime error (209) : no kernel image is available for execution on the device at /sda/okay/ai/project/ocrquesseg-svr/dyhead/csrc/cuda/SigmoidFocalLoss_cuda.cu:139

Hello,
Are masks necessary to work with this model or are bounding boxes enough?

I was trying to run the repo DynamicHead and trying to run the model with a custom dataset and have the following error. In the repo the launch function is called and I get these errors.

The command I run was

DETECTRON2_DATASETS=$DATASET python train_net.py --config configs/dyhead_r50_retina_fpn_1x.yaml --num-gpus 8

Logs

Traceback (most recent call last):
  File "train_net_custom.py", line 222, in <module>
    launch(
  File "/home/xxx/anaconda3/envs/detectron/lib/python3.8/site-packages/detectron2/engine/launch.py", line 82, in launch
    main_func(*args)
  File "train_net_custom.py", line 216, in main
    return trainer.train()
  File "/home/xxx/anaconda3/envs/detectron/lib/python3.8/site-packages/detectron2/engine/defaults.py", line 484, in train
    super().train(self.start_iter, self.max_iter)
  File "/home/xxx/anaconda3/envs/detectron/lib/python3.8/site-packages/detectron2/engine/train_loop.py", line 149, in train
    self.run_step()
  File "/home/xxx/anaconda3/envs/detectron/lib/python3.8/site-packages/detectron2/engine/defaults.py", line 494, in run_step
    self._trainer.run_step()
  File "/home/xxx/anaconda3/envs/detectron/lib/python3.8/site-packages/detectron2/engine/train_loop.py", line 294, in run_step
    self.optimizer.step()
  File "/home/xxx/anaconda3/envs/detectron/lib/python3.8/site-packages/torch/optim/lr_scheduler.py", line 65, in wrapper
    return wrapped(*args, **kwargs)
  File "/home/xxx/anaconda3/envs/detectron/lib/python3.8/site-packages/torch/optim/optimizer.py", line 89, in wrapper
    return func(*args, **kwargs)
  File "/home/xxx/anaconda3/envs/detectron/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 27, in decorate_context
    return func(*args, **kwargs)
  File "/home/xxx/anaconda3/envs/detectron/lib/python3.8/site-packages/torch/optim/sgd.py", line 110, in step
    F.sgd(params_with_grad,
  File "/home/xxx/anaconda3/envs/detectron/lib/python3.8/site-packages/torch/optim/_functional.py", line 160, in sgd
    d_p = d_p.add(param, alpha=weight_decay)
TypeError: add(): argument 'alpha' must be Number, not NoneType

Detectron2 version: 0.6

Expected behaviour is the training loop runs without error.

问一个问题已经很久没人回复了

Hello, Im very interested in your novel works.
Will you release the codes?

There are important files that Microsoft projects should all have that are not present in this repository. A pull request has been opened to add the missing file(s). When the pr is merged this issue will be closed automatically.

Microsoft teams can learn more about this effort and share feedback within the open source guidance available internally.

Merge this pull request

when running this line of code
spatial_output = spatial_layer(scale_output)

I got this error
~\AppData\Local\Continuum\anaconda3\lib\site-packages\torch\nn\modules\module.py in _call_impl(self, *input, **kwargs)
725 result = self._slow_forward(*input, **kwargs)
726 else:
--> 727 result = self.forward(*input, **kwargs)
728 for hook in itertools.chain(
729 _global_forward_hooks.values(),

TypeError: forward() takes 3 positional arguments but 4 were given

Traceback (most recent call last):
File "train_net.py", line 224, in
args=(args,),
File "/sda/program/anaconda3/envs/detectron2/lib/python3.7/site-packages/detectron2/engine/launch.py", line 82, in launch
main_func(*args)
File "train_net.py", line 210, in main
return trainer.train()
File "/sda/program/anaconda3/envs/detectron2/lib/python3.7/site-packages/detectron2/engine/defaults.py", line 484, in train
super().train(self.start_iter, self.max_iter)
File "/sda/program/anaconda3/envs/detectron2/lib/python3.7/site-packages/detectron2/engine/train_loop.py", line 149, in train
self.run_step()
File "/sda/program/anaconda3/envs/detectron2/lib/python3.7/site-packages/detectron2/engine/defaults.py", line 494, in run_step
self._trainer.run_step()
File "/sda/program/anaconda3/envs/detectron2/lib/python3.7/site-packages/detectron2/engine/train_loop.py", line 273, in run_step
loss_dict = self.model(data)
File "/sda/program/anaconda3/envs/detectron2/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/sda/okay/ai/project/ocrquesseg-svr/extra/atss.py", line 204, in forward
losses = self.losses(anchors, pred_logits, gt_labels, pred_anchor_deltas, gt_boxes, pred_centers)
File "/sda/okay/ai/project/ocrquesseg-svr/extra/atss.py", line 236, in losses
cls_loss = self.classification_loss_func(box_cls_flatten, labels_flatten.int()) / num_pos_avg_per_gpu
File "/sda/program/anaconda3/envs/detectron2/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/sda/okay/ai/project/ocrquesseg-svr/extra/sigmoid_focal_loss.py", line 48, in forward
loss = sigmoid_focal_loss_cuda(logits, targets, self.gamma, self.alpha)
File "/sda/okay/ai/project/ocrquesseg-svr/extra/sigmoid_focal_loss.py", line 20, in forward
logits, targets, num_classes, gamma, alpha
RuntimeError: cuda runtime error (209) : no kernel image is available for execution on the device at /sda/okay/ai/project/ocrquesseg-svr/dyhead/csrc/cuda/SigmoidFocalLoss_cuda.cu:139

I am trying to train resnet50-dyhead, but after 180000 steps, the loss is not reduce and there are not dectected instances in test images.
why? How to train dyhead?

Installing collected packages: dyhead
  Running setup.py develop for dyhead
    ERROR: Command errored out with exit status 1:
     command: /public/home/hpc2/anaconda3/envs/mask/bin/python -c 'import io, os, sys, setuptools, tokenize; sys.argv[0] = '"'"'/public/home/hpc2/rpc/DynamicHead/setup.py'"'"'; __file__='"'"'/public/home/hpc2/rpc/DynamicHead/setup.py'"'"';f = getattr(tokenize, '"'"'open'"'"', open)(__file__) if os.path.exists(__file__) else io.StringIO('"'"'from setuptools import setup; setup()'"'"');code = f.read().replace('"'"'\r\n'"'"', '"'"'\n'"'"');f.close();exec(compile(code, __file__, '"'"'exec'"'"'))' develop --no-deps
         cwd: /public/home/hpc2/rpc/DynamicHead/
    Complete output (71 lines):
    which: no nvcc in (/public/home/hpc2/anaconda3/envs/mask/bin:/public/home/hpc2/anaconda3/condabin:/usr/lib64/qt-3.3/bin:/public/home/hpc2/perl5/bin:/public/software/Modules/bin:/usr/local/bin:/usr/bin:/usr/local/sbin:/usr/sbin:/opt/ibutils/bin:/public/home/hpc2/.local/bin:/public/home/hpc204712181/bin)
    running develop
    running egg_info
    writing dyhead.egg-info/PKG-INFO
    writing dependency_links to dyhead.egg-info/dependency_links.txt
    writing top-level names to dyhead.egg-info/top_level.txt
    reading manifest file 'dyhead.egg-info/SOURCES.txt'
    writing manifest file 'dyhead.egg-info/SOURCES.txt'
    running build_ext
    building 'dyhead._C' extension
    gcc -pthread -B /public/home/hpc2/anaconda3/envs/mask/compiler_compat -Wl,--sysroot=/ -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -fPIC -I/public/home/hpc2/rpc/DynamicHead/dyhead/csrc -I/public/home/hpc2/anaconda3/envs/mask/lib/python3.7/site-packages/torch/include -I/public/home/hpc2/anaconda3/envs/mask/lib/python3.7/site-packages/torch/include/torch/csrc/api/include -I/public/home/hpc2/anaconda3/envs/mask/lib/python3.7/site-packages/torch/include/TH -I/public/home/hpc2/anaconda3/envs/mask/lib/python3.7/site-packages/torch/include/THC -I/public/home/hpc2/anaconda3/envs/mask/include/python3.7m -c /public/home/hpc2/rpc/DynamicHead/dyhead/csrc/vision.cpp -o build/temp.linux-x86_64-3.7/public/home/hpc204712181/rpc/DynamicHead/dyhead/csrc/vision.o -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=_C -D_GLIBCXX_USE_CXX11_ABI=0 -std=c++11
    cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++
    In file included from /public/home/hpc204712181/rpc/DynamicHead/dyhead/csrc/vision.cpp:2:0:
    /public/home/hpc2/rpc/DynamicHead/dyhead/csrc/deform_conv.h:10:5: error: ‘at’ has not been declared
         at::Tensor input,
         ^
    /public/home/hpc2/rpc/DynamicHead/dyhead/csrc/deform_conv.h:11:5: error: ‘at’ has not been declared
         at::Tensor weight,
         ^

error: command 'gcc' failed with exit status 1
Error seems that at::Tensor can't be recognized, how to solve this problem?
My gcc version is 5.4.0

In Table. 2
The sizes of models with <= 4 blocks are smaller than the baseline.
What modules are removed from the baseline? FPN?

We implement your Dynamichead with MaskRCNN-r50 on mmdetection. results are worse than maskrcnn, and it took almost twice as long.

Hi, thanks for your nice work.

I am wondering the performance of DynamicHead when the number of levels in the pyramid is one (non-pyramid structure). Thanks

Best,
Xinge

这是Dynamic head的代码吗？还是文章有造假的嫌疑？

Hello,
I would like to ask if anyone is facing the below issue:
TypeError: add(): argument 'alpha' must be Number, not NoneType

The steps I followed are:

1. python -m pip install -e DynamicHead.
2. Added custom dataset using register_coco_instance
3. Update config in def setup(args) with custom dataset name and the number of classes:
   cfg.DATASETS.TRAIN = ('coco_docbank_train',) cfg.MODEL.ROI_HEADS.NUM_CLASSES = 13
4. Run the model:
   python train_net.py --config configs/dyhead_r50_rcnn_fpn_1x.yaml --num-gpus 1.

The detailed error message goes like this:

File "train_net.py", line 204, in <module> 
    launch(
  File "/opt/conda/lib/python3.8/site-packages/detectron2/engine/launch.py", line 82, in launch
    main_func(*args)
  File "train_net.py", line 198, in main
    return trainer.train()
  File "/opt/conda/lib/python3.8/site-packages/detectron2/engine/defaults.py", line 484, in train
    super().train(self.start_iter, self.max_iter)
  File "/opt/conda/lib/python3.8/site-packages/detectron2/engine/train_loop.py", line 149, in train
    self.run_step()
  File "/opt/conda/lib/python3.8/site-packages/detectron2/engine/defaults.py", line 494, in run_step
    self._trainer.run_step()
  File "/opt/conda/lib/python3.8/site-packages/detectron2/engine/train_loop.py", line 294, in run_step
    self.optimizer.step()
  File "/opt/conda/lib/python3.8/site-packages/torch/optim/lr_scheduler.py", line 65, in wrapper
    return wrapped(*args, **kwargs)
  File "/opt/conda/lib/python3.8/site-packages/torch/optim/optimizer.py", line 88, in wrapper
    return func(*args, **kwargs)
  File "/opt/conda/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 28, in decorate_context
    return func(*args, **kwargs)
  File "/opt/conda/lib/python3.8/site-packages/torch/optim/sgd.py", line 136, in step
    F.sgd(params_with_grad,
  File "/opt/conda/lib/python3.8/site-packages/torch/optim/_functional.py", line 164, in sgd
    d_p = d_p.add(param, alpha=weight_decay)
TypeError: add(): argument 'alpha' must be Number, not NoneType

Environment Details:
sys.platform = linux
Python = 3.8.12
numpy = 1.21.2
detectron2 = 0.6
CUDA = 11.4
PyTorch = 1.10.0
torchvision = 0.11.0a0
fvcore = 0.1.5.post20211023
iopath = 0.1.9
cv2 = 4.5.4

Kindly request for assistance if you are aware of the solution.