Implementation of the Swin Transformer in PyTorch.

Overview

Linear Self Attention

Swin Transformer - PyTorch

Implementation of the Swin Transformer architecture. This paper presents a new vision Transformer, called Swin Transformer, that capably serves as a general-purpose backbone for computer vision. Challenges in adapting Transformer from language to vision arise from differences between the two domains, such as large variations in the scale of visual entities and the high resolution of pixels in images compared to words in text. To address these differences, we propose a hierarchical Transformer whose representation is computed with shifted windows. The shifted windowing scheme brings greater efficiency by limiting self-attention computation to non-overlapping local windows while also allowing for cross-window connection. This hierarchical architecture has the flexibility to model at various scales and has linear computational complexity with respect to image size. These qualities of Swin Transformer make it compatible with a broad range of vision tasks, including image classification (86.4 top-1 accuracy on ImageNet-1K) and dense prediction tasks such as object detection (58.7 box AP and 51.1 mask AP on COCO test-dev) and semantic segmentation (53.5 mIoU on ADE20K val). Its performance surpasses the previous state-of-the-art by a large margin of +2.7 box AP and +2.6 mask AP on COCO, and +3.2 mIoU on ADE20K, demonstrating the potential of Transformer-based models as vision backbones.

This is NOT the official repository of the Swin Transformer. At the moment in time the official code of the authors is not available yet but can be found later at: https://github.com/microsoft/Swin-Transformer.

All credits go to the authors Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin and Baining Guo.

Install

$ pip install swin-transformer-pytorch

or (if you clone the repository)

$ pip install -r requirements.txt

Usage

import torch
from swin_transformer_pytorch import SwinTransformer

net = SwinTransformer(
    hidden_dim=96,
    layers=(2, 2, 6, 2),
    heads=(3, 6, 12, 24),
    channels=3,
    num_classes=3,
    head_dim=32,
    window_size=7,
    downscaling_factors=(4, 2, 2, 2),
    relative_pos_embedding=True
)
dummy_x = torch.randn(1, 3, 224, 224)
logits = net(dummy_x)  # (1,3)
print(net)
print(logits)

Parameters

  • hidden_dim: int.
    What hidden dimension you want to use for the architecture, noted C in the original paper
  • layers: 4-tuple of ints divisible by 2.
    How many layers in each stage to apply. Every int should be divisible by two because we are always applying a regular and a shifted SwinBlock together.
  • heads: 4-tuple of ints
    How many heads in each stage to apply.
  • channels: int.
    Number of channels of the input.
  • num_classes: int.
    Num classes the output should have.
  • head_dim: int.
    What dimension each head should have.
  • window_size: int.
    What window size to use, make sure that after each downscaling the image dimensions are still divisible by the window size.
  • downscaling_factors: 4-tuple of ints.
    What downscaling factor to use in each stage. Make sure image dimension is large enough for downscaling factors.
  • relative_pos_embedding: bool.
    Whether to use learnable relative position embedding (2M-1)x(2M-1) or full positional embeddings (M²xM²).

TODO

  • Adjust code for and validate on ImageNet-1K and COCO 2017

References

Some part of the code is adapted from the PyTorch - VisionTransformer repository https://github.com/lucidrains/vit-pytorch , which provides a very clean VisionTransformer implementation to start with.

Citations

@misc{liu2021swin,
      title={Swin Transformer: Hierarchical Vision Transformer using Shifted Windows}, 
      author={Ze Liu and Yutong Lin and Yue Cao and Han Hu and Yixuan Wei and Zheng Zhang and Stephen Lin and Baining Guo},
      year={2021},
      eprint={2103.14030},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}
Comments
  • about widow-size

    about widow-size

    Dear Sir, Thank you very much for your great work. I would like to ask if you have any suggestions on how to set the window size. For 224x224 input, window size set to 7 is reasonable because it can divide by 7, but for other sizes, such as 768x768 in cityscapes, 7 will undoubtedly report an error since 768 / 32=24 , so it looks like the window setting is very subtle. The close value is 8, but is the window setting the same as the convolution kernel, where odd numbers work better? Also, is it possible to set different window sizes at different stages, which seems to be feasible for non-regular image sizes. Since the window size is a very critical hyperparameter that determines the perceptual field and the amount of computation, would like to request your opinion, thanks!

    opened by huixiancheng 9
  • relative pos embedding errs out with

    relative pos embedding errs out with "IndexError: tensors used as indices must be long, byte or bool tensors"

    Very big thanks for making this implementation! I just upgraded to the relative pos embedding update from an hour ago and in trying to train get this type error.

    ---> 32         y_pred = model(images)
         33         #print(f" y_pred = {y_pred}")
         34         #print(f" y_pred shape = {y_pred.shape}")
    
    ~\anaconda3\envs\fastai2\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(),
    
    ~\cdetr\cdetr_utils\transformer\swin_transformer.py in forward(self, img)
        229 
        230     def forward(self, img):
    --> 231         x = self.stage1(img)
        232         x = self.stage2(x)
        233         x = self.stage3(x)
    
    ~\anaconda3\envs\fastai2\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(),
    
    ~\cdetr\cdetr_utils\transformer\swin_transformer.py in forward(self, x)
        189         x = self.patch_partition(x)
        190         for regular_block, shifted_block in self.layers:
    --> 191             x = regular_block(x)
        192             x = shifted_block(x)
        193         return x.permute(0, 3, 1, 2)
    
    ~\anaconda3\envs\fastai2\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(),
    
    ~\cdetr\cdetr_utils\transformer\swin_transformer.py in forward(self, x)
        148 
        149     def forward(self, x):
    --> 150         x = self.attention_block(x)
        151         x = self.mlp_block(x)
        152         return x
    
    ~\anaconda3\envs\fastai2\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(),
    
    ~\cdetr\cdetr_utils\transformer\swin_transformer.py in forward(self, x, **kwargs)
         21 
         22     def forward(self, x, **kwargs):
    ---> 23         return self.fn(x, **kwargs) + x
         24 
         25 
    
    ~\anaconda3\envs\fastai2\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(),
    
    ~\cdetr\cdetr_utils\transformer\swin_transformer.py in forward(self, x, **kwargs)
         31 
         32     def forward(self, x, **kwargs):
    ---> 33         return self.fn(self.norm(x), **kwargs)
         34 
         35 
    
    ~\anaconda3\envs\fastai2\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(),
    
    ~\cdetr\cdetr_utils\transformer\swin_transformer.py in forward(self, x)
        116 
        117         if self.relative_pos_embedding:
    --> 118             dots += self.pos_embedding[self.relative_indices[:, :, 0], self.relative_indices[:, :, 1]]
        119         else:
        120             dots += self.pos_embedding
    
    IndexError: tensors used as indices must be long, byte or bool tensors
    
    opened by lessw2020 8
  • fail to run the code

    fail to run the code

    Hi, i'm intereted in your code! But when i run the example of it,

    Traceback (most recent call last): File "D:/Code/Pytorch/swin-transformer-pytorch-0.4/example.py", line 16, in logits = net(dummy_x) # (1,3) File "D:\Softwares\Anaconda\envs\pytorch_18\lib\site-packages\torch\nn\modules\module.py", line 889, in _call_impl result = self.forward(*input, **kwargs) File "D:\Code\Pytorch\swin-transformer-pytorch-0.4\swin_transformer_pytorch\swin_transformer.py", line 219, in forward x = self.stage1(img) File "D:\Softwares\Anaconda\envs\pytorch_18\lib\site-packages\torch\nn\modules\module.py", line 889, in _call_impl result = self.forward(*input, **kwargs) File "D:\Code\Pytorch\swin-transformer-pytorch-0.4\swin_transformer_pytorch\swin_transformer.py", line 190, in forward x = regular_block(x) File "D:\Softwares\Anaconda\envs\pytorch_18\lib\site-packages\torch\nn\modules\module.py", line 889, in _call_impl result = self.forward(*input, **kwargs) File "D:\Code\Pytorch\swin-transformer-pytorch-0.4\swin_transformer_pytorch\swin_transformer.py", line 149, in forward x = self.attention_block(x) File "D:\Softwares\Anaconda\envs\pytorch_18\lib\site-packages\torch\nn\modules\module.py", line 889, in _call_impl result = self.forward(*input, **kwargs) File "D:\Code\Pytorch\swin-transformer-pytorch-0.4\swin_transformer_pytorch\swin_transformer.py", line 22, in forward return self.fn(x, **kwargs) + x File "D:\Softwares\Anaconda\envs\pytorch_18\lib\site-packages\torch\nn\modules\module.py", line 889, in _call_impl result = self.forward(*input, **kwargs) File "D:\Code\Pytorch\swin-transformer-pytorch-0.4\swin_transformer_pytorch\swin_transformer.py", line 32, in forward return self.fn(self.norm(x), **kwargs) File "D:\Softwares\Anaconda\envs\pytorch_18\lib\site-packages\torch\nn\modules\module.py", line 889, in _call_impl result = self.forward(*input, **kwargs) File "D:\Code\Pytorch\swin-transformer-pytorch-0.4\swin_transformer_pytorch\swin_transformer.py", line 117, in forward dots += self.pos_embedding[self.relative_indices[:, :, 0], self.relative_indices[:, :, 1]] IndexError: tensors used as indices must be long, byte or bool tensors

    And when i change the type to long, the code has another error.

    Traceback (most recent call last): File "D:/Code/Pytorch/swin-transformer-pytorch-0.4/example.py", line 16, in logits = net(dummy_x) # (1,3) File "D:\Softwares\Anaconda\envs\pytorch_18\lib\site-packages\torch\nn\modules\module.py", line 889, in _call_impl result = self.forward(*input, **kwargs) File "D:\Code\Pytorch\swin-transformer-pytorch-0.4\swin_transformer_pytorch\swin_transformer.py", line 219, in forward x = self.stage1(img) File "D:\Softwares\Anaconda\envs\pytorch_18\lib\site-packages\torch\nn\modules\module.py", line 889, in _call_impl result = self.forward(*input, **kwargs) File "D:\Code\Pytorch\swin-transformer-pytorch-0.4\swin_transformer_pytorch\swin_transformer.py", line 188, in forward x = self.patch_partition(x) File "D:\Softwares\Anaconda\envs\pytorch_18\lib\site-packages\torch\nn\modules\module.py", line 889, in _call_impl result = self.forward(*input, **kwargs) File "D:\Code\Pytorch\swin-transformer-pytorch-0.4\swin_transformer_pytorch\swin_transformer.py", line 164, in forward x = self.patch_merge(x).view(b, -1, new_h, new_w).permute(0, 2, 3, 1) File "D:\Softwares\Anaconda\envs\pytorch_18\lib\site-packages\torch\nn\modules\module.py", line 889, in _call_impl result = self.forward(*input, **kwargs) File "D:\Softwares\Anaconda\envs\pytorch_18\lib\site-packages\torch\nn\modules\fold.py", line 295, in forward self.padding, self.stride) File "D:\Softwares\Anaconda\envs\pytorch_18\lib\site-packages\torch\nn\functional.py", line 4313, in unfold return torch._C._nn.im2col(input, _pair(kernel_size), _pair(dilation), _pair(padding), _pair(stride)) RuntimeError: "im2col_out_cpu" not implemented for 'Long'

    opened by ShujinW 5
  • Cyclic shift with masking

    Cyclic shift with masking

    Hello sir, I'm trying to understand "efficient batch computation" which the authors suggested. Probably because of my short knowledge, it was hard to get how it works. Your implementation really helped me for understanding its mechanism, thanks a lot!

    Here's my question, it seems the masked area of q * k / sqrt(d) vanishes during the computation of self-attention. I'm not sure that I understood the code correctly, but is this originally intended in the paper? I'm wondering if each subwindow's self-attention might be computed before reversing. image

    Apology if I misunderstood something, and thanks again!

    opened by Hayoung93 4
  • Shifting attention-calculating windows

    Shifting attention-calculating windows

    Hello, sir. A question popped up again, unfortunately.

    I've followed your shifting code, and it seems to have a difference with (my comprehension of) the paper. I understood the behavior of the original paper's window shifting as a black arrow in the image below (self-attention is calculated with elements inside of bold lines). The left red arrow points to the result of patch-wise rolling and the right red arrow points results of rolling the entire feature map. In my opinion, self-attention should be computed according to the right-top figure, therefore, boxes of right-bottom should be used (green dot-line separates subwindows) which each region in the right-top figure preserves.

    Please let me know if I misunderstood your code or something in the paper. Thanks a lot!

    Additionally, this is how I mimicked your code:

    import torch
    from einops import rearrange
    A = torch.Tensor(list(range(1, 17))).view(1, 4, 4)
    A_patched = A.view(4, 2, 2).permute(1, 2, 0).view(1, 2, 2, 4)
    A_patched_rolled = torch.roll(A_patched, shifts=(-1, -1), dims=(1, 2))
    A_rearranged = rearrange(A, 'a (b c) (d e)->a (b d) (c e)', b=2, d=2)
    A_rearranged_rolled = torch.roll(A_rearranged, shifts=(-1, -1), dims=(1, 2))
    A_rearranged_rolled2 = torch.roll(A_rearranged, shifts=(1, 1), dims=(1, 2))
    

    where A can be considered as a 4x4 feature map (though element order is not matched with image above), A_patched is a divided version of A, and A_patched_rolled is patch-wise shifted version of A_patched, following torch.roll(x, shifts=(self.displacement, self.displacement), dims=(1, 2)) in your code. A_rearranged is rearranged to match the image above.

    <---A_patched<---A_patched_rolled

    >>> A
    tensor([[[ 1.,  2.,  3.,  4.],
             [ 5.,  6.,  7.,  8.],
             [ 9., 10., 11., 12.],
             [13., 14., 15., 16.]]])
    >>> A_patched
    tensor([[[[ 1.,  5.,  9., 13.],
              [ 2.,  6., 10., 14.]],
    
             [[ 3.,  7., 11., 15.],
              [ 4.,  8., 12., 16.]]]])
    >>> A_patched_rolled
    tensor([[[[ 4.,  8., 12., 16.],
              [ 3.,  7., 11., 15.]],
    
             [[ 2.,  6., 10., 14.],
              [ 1.,  5.,  9., 13.]]]])
    >>> A_rearranged
    tensor([[[ 1.,  2.,  5.,  6.],
             [ 3.,  4.,  7.,  8.],
             [ 9., 10., 13., 14.],
             [11., 12., 15., 16.]]])
    >>> A_rearranged_rolled
    tensor([[[ 4.,  7.,  8.,  3.],
             [10., 13., 14.,  9.],
             [12., 15., 16., 11.],
             [ 2.,  5.,  6.,  1.]]])
    >>> A_rearranged_rolled2
    tensor([[[16., 11., 12., 15.],
             [ 6.,  1.,  2.,  5.],
             [ 8.,  3.,  4.,  7.],
             [14.,  9., 10., 13.]]])
    
    opened by Hayoung93 2
  • How to use for generation work

    How to use for generation work

    Thanks for your great work. I do the task of image generation. In my opinion, the current swin-transformer is an encode structure. Is there a corresponding swin-transformer that can be used for decode?

    opened by yinyiyu 2
  • Runtime error

    Runtime error

    I'm running an error in your code at line 117 dots += self.pos_embedding[self.relative_indices[:, :, 0], self.relative_indices[:, :, 1]] IndexError: tensors used as indices must be long, byte or bool tensors

    opened by QinchengZhang 1
  • A question about qk_scale

    A question about qk_scale

    Hello @berniwal , I have a question about this: https://github.com/berniwal/swin-transformer-pytorch/blob/c921ebf914c6ea9734bb260ada395e3746c85402/swin_transformer_pytorch/swin_transformer.py#L76

    what's the function of the scale?I can't understand why do this.

    Best regards

    opened by Sample-design-alt 0
  • Issues related to patch merging implementation

    Issues related to patch merging implementation

    In this repository, patch merging is implemented with nn.Unfold, but it is expected to behave differently than the official implementation.

    https://github.com/microsoft/Swin-Transformer/blob/6ded2577413b68cbbd89f08391465788ed73030e/models/swin_transformer.py#L291

    Is there something I'm missing out on?

    opened by lee-gwang 1
  • why the createmask function is 49*49?

    why the createmask function is 49*49?

    def create_mask(window_size, displacement, upper_lower, left_right): mask = torch.zeros(window_size ** 2, window_size ** 2)

    it is 49*49 in all tne swin network,why?

    opened by henbucuoshanghai 1
  • apply to other dataset

    apply to other dataset

    hello,thanks for the work you had done very much and i have a question that how can i apply this code to train a vit model on other dataset,how can i to adjust those parameters?

    opened by jieweilai 0
  • deeplabv3 + swintransformer

    deeplabv3 + swintransformer

    i try this swintransformer on deeplabv3 (https://github.com/VainF/DeepLabV3Plus-Pytorch), errors are found:

    Exception has occurred: EinopsError Error while processing rearrange-reduction pattern "b (nw_h w_h) (nw_w w_w) (h d) -> b h (nw_h nw_w) (w_h w_w) d". Input tensor shape: torch.Size([1, 104, 104, 96]). Additional info: {'h': 3, 'w_h': 7, 'w_w': 7}. Shape mismatch, can't divide axis of length 104 in chunks of 7

    During handling of the above exception, another exception occurred:

    File "D:\TangYong\Src\VS\Python\PyTorch\deeplabv3-vainf\network\backbone\swintransformer.py", line 111, in lambda t: rearrange(t, 'b (nw_h w_h) (nw_w w_w) (h d) -> b h (nw_h nw_w) (w_h w_w) d', File "D:\TangYong\Src\VS\Python\PyTorch\deeplabv3-vainf\network\backbone\swintransformer.py", line 110, in forward q, k, v = map( File "D:\TangYong\Src\VS\Python\PyTorch\deeplabv3-vainf\network\backbone\swintransformer.py", line 32, in forward return self.fn(self.norm(x), **kwargs) File "D:\TangYong\Src\VS\Python\PyTorch\deeplabv3-vainf\network\backbone\swintransformer.py", line 22, in forward return self.fn(x, **kwargs) + x File "D:\TangYong\Src\VS\Python\PyTorch\deeplabv3-vainf\network\backbone\swintransformer.py", line 149, in forward

    thank you for your answer.

    opened by TangYong1975 1
Releases(0.4)
VarCLR: Variable Semantic Representation Pre-training via Contrastive Learning

    VarCLR: Variable Representation Pre-training via Contrastive Learning New: Paper accepted by ICSE 2022. Preprint at arXiv! This repository contain

squaresLab 32 Oct 24, 2022
Official code for our ICCV paper: "From Continuity to Editability: Inverting GANs with Consecutive Images"

GANInversion_with_ConsecutiveImgs Official code for our ICCV paper: "From Continuity to Editability: Inverting GANs with Consecutive Images" https://a

QingyangXu 38 Dec 07, 2022
HAT: Hierarchical Aggregation Transformers for Person Re-identification

HAT: Hierarchical Aggregation Transformers for Person Re-identification

11 Sep 05, 2022
ZEBRA: Zero Evidence Biometric Recognition Assessment

ZEBRA: Zero Evidence Biometric Recognition Assessment license: LGPLv3 - please reference our paper version: 2020-06-11 author: Andreas Nautsch (EURECO

Voice Privacy Challenge 2 Dec 12, 2021
Deep Q-Learning Network in pytorch (not actively maintained)

pytoch-dqn This project is pytorch implementation of Human-level control through deep reinforcement learning and I also plan to implement the followin

Hung-Tu Chen 342 Jan 01, 2023
Solution to the Weather4cast 2021 challenge

This code was used for the entry by the team "antfugue" for the Weather4cast 2021 Challenge. Below, you can find the instructions for generating predi

Jussi Leinonen 13 Jan 03, 2023
This repository contains the database and code used in the paper Embedding Arithmetic for Text-driven Image Transformation

This repository contains the database and code used in the paper Embedding Arithmetic for Text-driven Image Transformation (Guillaume Couairon, Holger

Meta Research 31 Oct 17, 2022
This is the pytorch implementation for the paper: *Learning Accurate Performance Predictors for Ultrafast Automated Model Compression*, which is in submission to TPAMI

SeerNet This is the pytorch implementation for the paper: Learning Accurate Performance Predictors for Ultrafast Automated Model Compression, which is

3 May 01, 2022
Athena is the only tool that you will ever need to optimize your portfolio.

Athena Portfolio optimization is the process of selecting the best portfolio (asset distribution), out of the set of all portfolios being considered,

Indrajit 1 Mar 25, 2022
Official code of ICCV2021 paper "Residual Attention: A Simple but Effective Method for Multi-Label Recognition"

CSRA This is the official code of ICCV 2021 paper: Residual Attention: A Simple But Effective Method for Multi-Label Recoginition Demo, Train and Vali

163 Dec 22, 2022
Implementation of TimeSformer, a pure attention-based solution for video classification

TimeSformer - Pytorch Implementation of TimeSformer, a pure and simple attention-based solution for reaching SOTA on video classification.

Phil Wang 602 Jan 03, 2023
Axel - 3D printed robotic hands and they controll with Raspberry Pi and Arduino combo

Axel It's our graduation project about 3D printed robotic hands and they control

0 Feb 14, 2022
Near-Optimal Sparse Allreduce for Distributed Deep Learning (published in PPoPP'22)

Near-Optimal Sparse Allreduce for Distributed Deep Learning (published in PPoPP'22) Ok-Topk is a scheme for distributed training with sparse gradients

Shigang Li 9 Oct 29, 2022
Direct LiDAR Odometry: Fast Localization with Dense Point Clouds

Direct LiDAR Odometry: Fast Localization with Dense Point Clouds DLO is a lightweight and computationally-efficient frontend LiDAR odometry solution w

VECTR at UCLA 369 Dec 30, 2022
A Fast and Accurate One-Stage Approach to Visual Grounding, ICCV 2019 (Oral)

One-Stage Visual Grounding ***** New: Our recent work on One-stage VG is available at ReSC.***** A Fast and Accurate One-Stage Approach to Visual Grou

Zhengyuan Yang 118 Dec 05, 2022
[NeurIPS 2021] Low-Rank Subspaces in GANs

Low-Rank Subspaces in GANs Figure: Image editing results using LowRankGAN on StyleGAN2 (first three columns) and BigGAN (last column). Low-Rank Subspa

112 Dec 28, 2022
[TOG 2021] PyTorch implementation for the paper: SofGAN: A Portrait Image Generator with Dynamic Styling.

This repository contains the official PyTorch implementation for the paper: SofGAN: A Portrait Image Generator with Dynamic Styling. We propose a SofGAN image generator to decouple the latent space o

Anpei Chen 694 Dec 23, 2022
[ICRA 2022] An opensource framework for cooperative detection. Official implementation for OPV2V.

OpenCOOD OpenCOOD is an Open COOperative Detection framework for autonomous driving. It is also the official implementation of the ICRA 2022 paper OPV

Runsheng Xu 322 Dec 23, 2022
[ICCV 2021] Deep Hough Voting for Robust Global Registration

Deep Hough Voting for Robust Global Registration, ICCV, 2021 Project Page | Paper | Video Deep Hough Voting for Robust Global Registration Junha Lee1,

57 Nov 28, 2022
Deep Learning: Architectures & Methods Project: Deep Learning for Audio Super-Resolution

Deep Learning: Architectures & Methods Project: Deep Learning for Audio Super-Resolution Figure: Example visualization of the method and baseline as a

Oliver Hahn 16 Dec 23, 2022