This is an official implementation of the paper "Distance-aware Quantization", accepted to ICCV2021.

CV Lab @ Yonsei University

Last update: Nov 4, 2022

Related tags

Deep Learning DAQ

Overview

PyTorch implementation of DAQ

This is an official implementation of the paper "Distance-aware Quantization", accepted to ICCV2021.

For more information, checkout the project site [website].

Getting started

Dependencies

Python 3.6
PyTorch = 1.5.0

Datasets

Cifar-10
- This can be automatically downloaded by learning our code
ImageNet
- This is available at here

Training & Evaluation

First, clone our github repository.

$ git clone https://github.com/cvlab-yonsei/DAQ.git

Cifar-10 dataset (ResNet-20 architecture)

First, download full-precision model into results/ folder. Link: [weights]
Note that you create results/ directory manually.

# Cifar-10 & ResNet-20 W1A1 model
$ python cifar10_train.py --config configs/DAQ/resnet20_DAQ_W1A1.yml
# Cifar-10 & ResNet-20 W1A32 model
$ python cifar10_train.py --config configs/DAQ/resnet20_DAQ_W1A32.yml

ImageNet dataset (ResNet-18 architecture)

Will be released

# ImageNet & ResNet-18 W1A1 model
# ImageNet & ResNet-18 W1A1 model

Using the pretrained models

ResNet-20
- W1A1 [weights]
- W1A32 [weights]
ResNet-18
- Will be released

Citation

@inproceedings{kim2021daq,
    author={Kim, Dohyung  and Lee, Junghyup and Ham, Bumsub},
    title={Distance-aware Quantization},
    booktitle={Proceedings of International Conference on Computer Vision},
    year={2021},
}

Credit

ImageNet training code: [PyTorch official example code]
ResNet-18 models: [PyTorch official code]
ResNet-20 model: [ResNet on CIFAR10] [IRNet]
Quantized modules: [DSQ]

You might also like...

Official implementation of NLOS-OT: Passive Non-Line-of-Sight Imaging Using Optimal Transport (IEEE TIP, accepted)

NLOS-OT Official implementation of NLOS-OT: Passive Non-Line-of-Sight Imaging Using Optimal Transport (IEEE TIP, accepted) Description In this reposit

16 Dec 16, 2022

This repository provides the official implementation of 'Learning to ignore: rethinking attention in CNNs' accepted in BMVC 2021.

inverse_attention This repository provides the official implementation of 'Learning to ignore: rethinking attention in CNNs' accepted in BMVC 2021. Le

5 Jul 8, 2022

This is the official implementation code repository of Underwater Light Field Retention : Neural Rendering for Underwater Imaging (Accepted by CVPR Workshop2022 NTIRE)

Underwater Light Field Retention : Neural Rendering for Underwater Imaging (UWNR) (Accepted by CVPR Workshop2022 NTIRE) Authors: Tian Ye†, Sixiang Che

17 Dec 14, 2022

Official PyTorch Implementation of Rank & Sort Loss [ICCV2021]

PyTorch implementation of our ICCV2021 paper: StructDepth: Leveraging the structural regularities for self-supervised indoor depth estimation

StructDepth PyTorch implementation of our ICCV2021 paper: StructDepth: Leveraging the structural regularities for self-supervised indoor depth estimat

112 Nov 28, 2022

Comments

Gussian kernel and Rescale f(y)
Thanks for your great paper, I have two questions about the code

The first is the 1-dimsion Gussion kernel, in the code, is it k_p = torch.exp(-(torch.pow(self.q_value-max_idx, 2).float()/(sigma**2)))? all the elements in k_p is the same

The second is the Rescale f(y)，Is there any more math explanation? the paper use q_c and q_f, while the code use (x-q_c) and (x-q_f), there still exits a gap for me.

q_var = self.q_value.clone() q_var[0] = q_var[0] - (1/(torch.exp(torch.tensor(T_ori).float()) - 1)) q_var[1] = q_var[1] + (1/(torch.exp(torch.tensor(T_ori).float()) - 1)) output = (q_var * prob).sum(dim=0) output = output + x_floor
opened by qiulinzhang 5
abs backward function

Thanks for your great paper. I have one question about the class absol, line 31, https://github.com/cvlab-yonsei/DAQ/blob/main/models/resnet20_DAQ.py, I don't understand your backward function, there seems a lot of redundant operation after grad_input = torch.sign(input). Is there any specific explanation about this backward function.

opened by qiulinzhang 3
Eq. 7 in the paper

Hi!

When reading the paper, I calculated that the closer to the transition point the smaller \beta^* is in Eq. 7. Because one part of the denominator is very large and the other part tends to 0, which contradicts the description in your paper.

Thanks.

opened by BobxmuMa 0
Paper formula is not the same as code

Hi，thanks for your work，When I read the paper formula and code, I found a little different, like where λ = 1/(exp(γ) + 1) as stated in Sec. 3.3，paper is written as +1, but the code is - 1, why?

https://github.com/cvlab-yonsei/DAQ/blob/e8e9af4cc9ebef2b77800fc3b04ecbf722205dab/models/resnet20_DAQ.py#L133

opened by aksenventwo 0

This is an official implementation of the paper "Distance-aware Quantization", accepted to ICCV2021.

Related tags

Overview

PyTorch implementation of DAQ

Getting started

Dependencies

Datasets

Training & Evaluation

Using the pretrained models

Citation

Credit

You might also like...

Official implementation of NLOS-OT: Passive Non-Line-of-Sight Imaging Using Optimal Transport (IEEE TIP, accepted)

This repository provides the official implementation of 'Learning to ignore: rethinking attention in CNNs' accepted in BMVC 2021.

This is the official implementation code repository of Underwater Light Field Retention : Neural Rendering for Underwater Imaging (Accepted by CVPR Workshop2022 NTIRE)

Official PyTorch Implementation of Rank & Sort Loss [ICCV2021]

Official implementation of "A Unified Objective for Novel Class Discovery", ICCV2021 (Oral)

Implementation of accepted AAAI 2021 paper: Deep Unsupervised Image Hashing by Maximizing Bit Entropy

Implementation of ICCV2021(Oral) paper - VMNet: Voxel-Mesh Network for Geodesic-aware 3D Semantic Segmentation

Implementation for our ICCV2021 paper: Internal Video Inpainting by Implicit Long-range Propagation

PyTorch implementation of our ICCV2021 paper: StructDepth: Leveraging the structural regularities for self-supervised indoor depth estimation

Comments

Gussian kernel and Rescale f(y)

abs backward function

Eq. 7 in the paper

Paper formula is not the same as code

Owner

CV Lab @ Yonsei University

This project is the official implementation of our accepted ICLR 2021 paper BiPointNet: Binary Neural Network for Point Clouds.

Official PyTorch implementation of the preprint paper "Stylized Neural Painting", accepted to CVPR 2021.

Official PyTorch implementation of the paper "Recycling Discriminator: Towards Opinion-Unaware Image Quality Assessment Using Wasserstein GAN", accepted to ACM MM 2021 BNI Track.

This repository contains the official implementation code of the paper Improving Multimodal Fusion with Hierarchical Mutual Information Maximization for Multimodal Sentiment Analysis, accepted at EMNLP 2021.

The official implementation of paper Siamese Transformer Pyramid Networks for Real-Time UAV Tracking, accepted by WACV22

An official source code for paper Deep Graph Clustering via Dual Correlation Reduction, accepted by AAAI 2022

Official code of ICCV2021 paper "Residual Attention: A Simple but Effective Method for Multi-Label Recognition"

Official code for ICCV2021 paper "M3D-VTON: A Monocular-to-3D Virtual Try-on Network"

Official Repo for ICCV2021 Paper: Learning to Regress Bodies from Images using Differentiable Semantic Rendering

An official TensorFlow implementation of “CLCC: Contrastive Learning for Color Constancy” accepted at CVPR 2021.