Deep Residual Learning for Image Recognition

Overview

Deep Residual Learning for Image Recognition

This is a Torch implementation of "Deep Residual Learning for Image Recognition",Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun the winners of the 2015 ILSVRC and COCO challenges.

What's working: CIFAR converges, as per the paper.

What's not working yet: Imagenet. I also have only implemented Option (A) for the residual network bottleneck strategy.

Table of contents

Changes

  • 2016-02-01: Added others' preliminary results on ImageNet for the architecture. (I haven't found time to train ImageNet yet)
  • 2016-01-21: Completed the 'alternate solver' experiments on deep networks. These ones take quite a long time.
  • 2016-01-19:
    • New results: Re-ran the 'alternate building block' results on deeper networks. They have more of an effect.
    • Added a table of contents to avoid getting lost.
    • Added experimental artifacts (log of training loss and test error, the saved model, the any patches used on the source code, etc) for two of the more interesting experiments, for curious folks who want to reproduce our results. (These artifacts are hereby released under the zlib license.)
  • 2016-01-15:
    • New CIFAR results: I re-ran all the CIFAR experiments and updated the results. There were a few bugs: we were only testing on the first 2,000 images in the training set, and they were sampled with replacement. These new results are much more stable over time.
  • 2016-01-12: Release results of CIFAR experiments.

How to use

  • You need at least CUDA 7.0 and CuDNN v4.
  • Install Torch.
  • Install the Torch CUDNN V4 library: git clone https://github.com/soumith/cudnn.torch; cd cudnn; git co R4; luarocks make This will give you cudnn.SpatialBatchNormalization, which helps save quite a lot of memory.
  • Install nninit: luarocks install nninit.
  • Download CIFAR 10. Use --dataRoot to specify the location of the extracted CIFAR 10 folder.
  • Run train-cifar.lua.

CIFAR: Effect of model size

For this test, our goal is to reproduce Figure 6 from the original paper:

figure 6 from original paper

We train our model for 200 epochs (this is about 7.8e4 of their iterations on the above graph). Like their paper, we start at a learning rate of 0.1 and reduce it to 0.01 at 80 epochs and then to 0.01 at 160 epochs.

Training loss

Training loss curve

Testing error

Test error curve

Model My Test Error Reference Test Error from Tab. 6 Artifacts
Nsize=3, 20 layers 0.0829 0.0875 Model, Loss and Error logs, Source commit + patch
Nsize=5, 32 layers 0.0763 0.0751 Model, Loss and Error logs, Source commit + patch
Nsize=7, 44 layers 0.0714 0.0717 Model, Loss and Error logs, Source commit + patch
Nsize=9, 56 layers 0.0694 0.0697 Model, Loss and Error logs, Source commit + patch
Nsize=18, 110 layers, fancy policy¹ 0.0673 0.0661² Model, Loss and Error logs, Source commit + patch

We can reproduce the results from the paper to typically within 0.5%. In all cases except for the 32-layer network, we achieve very slightly improved performance, though this may just be noise.

¹: For this run, we started from a learning rate of 0.001 until the first 400 iterations. We then raised the learning rate to 0.1 and trained as usual. This is consistent with the actual paper's results.

²: Note that the paper reports the best run from five runs, as well as the mean. I consider the mean to be a valid test protocol, but I don't like reporting the 'best' score because this is effectively training on the test set. (This method of reporting effectively introduces an extra parameter into the model--which model to use from the ensemble--and this parameter is fitted to the test set)

CIFAR: Effect of model architecture

This experiment explores the effect of different NN architectures that alter the "Building Block" model inside the residual network.

The original paper used a "Building Block" similar to the "Reference" model on the left part of the figure below, with the standard convolution layer, batch normalization, and ReLU, followed by another convolution layer and batch normalization. The only interesting piece of this architecture is that they move the ReLU after the addition.

We investigated two alternate strategies.

Three different alternate CIFAR architectures

  • Alternate 1: Move batch normalization after the addition. (Middle) The reasoning behind this choice is to test whether normalizing the first term of the addition is desirable. It grew out of the mistaken belief that batch normalization always normalizes to have zero mean and unit variance. If this were true, building an identity building block would be impossible because the input to the addition always has unit variance. However, this is not true. BN layers have additional learnable scale and bias parameters, so the input to the batch normalization layer is not forced to have unit variance.

  • Alternate 2: Remove the second ReLU. The idea behind this was noticing that in the reference architecture, the input cannot proceed to the output without being modified by a ReLU. This makes identity connections technically impossible because negative numbers would always be clipped as they passed through the skip layers of the network. To avoid this, we could either move the ReLU before the addition or remove it completely. However, it is not correct to move the ReLU before the addition: such an architecture would ensure that the output would never decrease because the first addition term could never be negative. The other option is to simply remove the ReLU completely, sacrificing the nonlinear property of this layer. It is unclear which approach is better.

To test these strategies, we repeat the above protocol using the smallest (20-layer) residual network model.

(Note: The other experiments all use the leftmost "Reference" model.)

Training loss

Testing error

Architecture Test error
ReLU, BN before add (ORIG PAPER reimplementation) 0.0829
No ReLU, BN before add 0.0862
ReLU, BN after add 0.0834
No ReLU, BN after add 0.0823

All methods achieve accuracies within about 0.5% of each other. Removing ReLU and moving the batch normalization after the addition seems to make a small improvement on CIFAR, but there is too much noise in the test error curve to reliably tell a difference.

CIFAR: Effect of model architecture on deep networks

The above experiments on the 20-layer networks do not reveal any interesting differences. However, these differences become more pronounced when evaluated on very deep networks. We retry the above experiments on 110-layer (Nsize=19) networks.

Training loss

Testing error

Results:

  • For deep networks, it's best to put the batch normalization before the addition part of each building block layer. This effectively removes most of the batch normalization operations from the input skip paths. If a batch normalization comes after each building block, then there exists a path from the input straight to the output that passes through several batch normalizations in a row. This could be problematic because each BN is not idempotent (the effects of several BN layers accumulate).

  • Removing the ReLU layer at the end of each building block appears to give a small improvement (~0.6%)

Architecture Test error Artifacts
ReLU, BN before add (ORIG PAPER reimplementation) 0.0697 Model, Loss and Error logs, Source commit + patch
No ReLU, BN before add 0.0632 Model, Loss and Error logs, Source commit + patch
ReLU, BN after add 0.1356 Model, Loss and Error logs, Source commit + patch
No ReLU, BN after add 0.1230 Model, Loss and Error logs, Source commit + patch

ImageNet: Effect of model architecture (preliminary)

@ducha-aiki is performing preliminary experiments on imagenet. For ordinary CaffeNet networks, @ducha-aiki found that putting batch normalization after the ReLU layer may provide a small benefit compared to putting it before.

Second, results on CIFAR-10 often contradicts results on ImageNet. I.e., leaky ReLU > ReLU on CIFAR, but worse on ImageNet.

@ducha-aiki's more detailed results here: https://github.com/ducha-aiki/caffenet-benchmark/blob/master/batchnorm.md

CIFAR: Alternate training strategies (RMSPROP, Adagrad, Adadelta)

Can we improve on the basic SGD update rule with Nesterov momentum? This experiment aims to find out. Common wisdom suggests that alternate update rules may converge faster, at least initially, but they do not outperform well-tuned SGD in the long run.

Training loss curve

Testing error curve

In our experiments, vanilla SGD with Nesterov momentum and a learning rate of 0.1 eventually reaches the lowest test error. Interestingly, RMSPROP with learning rate 1e-2 achieves a lower training loss, but overfits.

Strategy Test error
Original paper: SGD + Nesterov momentum, 1e-1 0.0829
RMSprop, learrning rate = 1e-4 0.1677
RMSprop, 1e-3 0.1055
RMSprop, 1e-2 0.0945
Adadelta¹, rho = 0.3 0.1093
Adagrad, 1e-3 0.3536
Adagrad, 1e-2 0.1603
Adagrad, 1e-1 0.1255

¹: Adadelta does not use a learning rate, so we did not use the same learning rate policy as in the paper. We just let it run until convergence.

See Andrej Karpathy's CS231N notes for more details on each of these learning strategies.

CIFAR: Alternate training strategies on deep networks

Deeper networks are more prone to overfitting. Unlike the earlier experiments, all of these models (except Adagrad with a learning rate of 1e-3) achieve a loss under 0.1, but test error varies quite wildly. Once again, using vanilla SGD with Nesterov momentum achieves the lowest error.

Training loss

Testing error

Solver Testing error
Nsize=18, Original paper: Nesterov, 1e-1 0.0697
Nsize=18, RMSprop, 1e-4 0.1482
Nsize=18, RMSprop, 1e-3 0.0821
Nsize=18, RMSprop, 1e-2 0.0768
Nsize=18, RMSprop, 1e-1 0.1098
Nsize=18, Adadelta 0.0888
Nsize=18, Adagrad, 1e-3 0.3022
Nsize=18, Adagrad, 1e-2 0.1321
Nsize=18, Adagrad, 1e-1 0.1145

Effect of batch norm momentum

For our experiments, we use batch normalization using an exponential running mean and standard deviation with a momentum of 0.1, meaning that the running mean and std changes by 10% of its value at each batch. A value of 1.0 would cause the batch normalization layer to calculate the mean and standard deviation across only the current batch, and a value of 0 would cause the batch normalization layer to stop accumulating changes in the running mean and standard deviation.

The strictest interpretation of the original batch normalization paper is to calculate the mean and standard deviation across the entire training set at every update. This takes too long in practice, so the exponential average is usually used instead.

We attempt to see whether batch normalization momentum affects anything. We try different values away from the default, along with a "dynamic" update strategy that sets the momentum to 1 / (1+n), where n is the number of batches seen so far (N resets to 0 at every epoch). At the end of training for a certain epoch, this means the batch normalization's running mean and standard deviation is effectively calculated over the entire training set.

None of these effects appear to make a significant difference.

Test error curve

Strategy Test Error
BN, momentum = 1 just for fun 0.0863
BN, momentum = 0.01 0.0835
Original paper: BN momentum = 0.1 0.0829
Dynamic, reset every epoch. 0.0822

TODO: Imagenet

Owner
Kimmy
Kimmy
Application of K-means algorithm on a music dataset after a dimensionality reduction with PCA

PCA for dimensionality reduction combined with Kmeans Goal The Goal of this notebook is to apply a dimensionality reduction on a big dataset in order

Arturo Ghinassi 0 Sep 17, 2022
BasicVSR++: Improving Video Super-Resolution with Enhanced Propagation and Alignment

BasicVSR++: Improving Video Super-Resolution with Enhanced Propagation and Alignment

Holy Wu 35 Jan 01, 2023
Pytorch implementation of PTNet for high-resolution and longitudinal infant MRI synthesis

Pyramid Transformer Net (PTNet) Project | Paper Pytorch implementation of PTNet for high-resolution and longitudinal infant MRI synthesis. PTNet: A Hi

Xuzhe Johnny Zhang 6 Jun 08, 2022
simple demo codes for Learning to Teach with Dynamic Loss Functions

Learning to Teach with Dynamic Loss Functions This repo contains the simple demo for the NeurIPS-18 paper: Learning to Teach with Dynamic Loss Functio

Lijun Wu 15 Dec 30, 2021
Colossal-AI: A Unified Deep Learning System for Large-Scale Parallel Training

ColossalAI An integrated large-scale model training system with efficient parallelization techniques. arXiv: Colossal-AI: A Unified Deep Learning Syst

HPC-AI Tech 7.9k Jan 08, 2023
Bayesian Deep Learning and Deep Reinforcement Learning for Object Shape Error Response and Correction of Manufacturing Systems

Bayesian Deep Learning for Manufacturing 2.0 (dlmfg) Object Shape Error Response (OSER) Digital Lifecycle Management - In Process Quality Improvement

Sumit Sinha 30 Oct 31, 2022
🔥🔥High-Performance Face Recognition Library on PaddlePaddle & PyTorch🔥🔥

face.evoLVe: High-Performance Face Recognition Library based on PaddlePaddle & PyTorch Evolve to be more comprehensive, effective and efficient for fa

Zhao Jian 3.1k Jan 02, 2023
TRIQ implementation

TRIQ Implementation TF-Keras implementation of TRIQ as described in Transformer for Image Quality Assessment. Installation Clone this repository. Inst

Junyong You 115 Dec 30, 2022
交互式标注软件,暂定名 iann

iann 交互式标注软件,暂定名iann。 安装 按照官网介绍安装paddle。 安装其他依赖 pip install -r requirements.txt 运行 git clone https://github.com/PaddleCV-SIG/iann/ cd iann python iann

294 Dec 30, 2022
Mind the Trade-off: Debiasing NLU Models without Degrading the In-distribution Performance

Models for natural language understanding (NLU) tasks often rely on the idiosyncratic biases of the dataset, which make them brittle against test cases outside the training distribution.

Ubiquitous Knowledge Processing Lab 22 Jan 02, 2023
This is the pytorch implementation for the paper: Generalizable Mixed-Precision Quantization via Attribution Rank Preservation, which is accepted to ICCV2021.

GMPQ: Generalizable Mixed-Precision Quantization via Attribution Rank Preservation This is the pytorch implementation for the paper: Generalizable Mix

18 Sep 02, 2022
The Video-based Accident Detection System built in Python

Accident-detection-system About the Project This Repository contains the Video-based Accident Detection System built in Python. Contributors Yukta Gop

SURYAVANSHI SNEHAL BALKRISHNA 50 Dec 07, 2022
PaddlePaddle GAN library, including lots of interesting applications like First-Order motion transfer, wav2lip, picture repair, image editing, photo2cartoon, image style transfer, and so on.

English | 简体中文 PaddleGAN PaddleGAN provides developers with high-performance implementation of classic and SOTA Generative Adversarial Networks, and s

6.4k Jan 09, 2023
The official code repo of "HTS-AT: A Hierarchical Token-Semantic Audio Transformer for Sound Classification and Detection"

Hierarchical Token Semantic Audio Transformer Introduction The Code Repository for "HTS-AT: A Hierarchical Token-Semantic Audio Transformer for Sound

Knut(Ke) Chen 134 Jan 01, 2023
Perform zero-order Hankel Transform for an 1D array (float or real valued).

perform zero-order Hankel Transform for an 1D array (float or real valued). An discrete form of Parseval theorem is guaranteed. Suit for iterative problems.

1 Jan 17, 2022
AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty

AugMix Introduction We propose AugMix, a data processing technique that mixes augmented images and enforces consistent embeddings of the augmented ima

Google Research 876 Dec 17, 2022
[Link]mareteutral - pars tradg wth M []

pairs-trading-with-ML Jonathan Larkin, August 2017 One popular strategy classification is Pairs Trading. Though this category of strategies can exhibi

Jonathan Larkin 134 Jan 06, 2023
IAST: Instance Adaptive Self-training for Unsupervised Domain Adaptation (ECCV 2020)

This repo is the official implementation of our paper "Instance Adaptive Self-training for Unsupervised Domain Adaptation". The purpose of this repo is to better communicate with you and respond to y

CVSM Group - email: <a href=[email protected]"> 84 Dec 12, 2022
An official implementation of "SFNet: Learning Object-aware Semantic Correspondence" (CVPR 2019, TPAMI 2020) in PyTorch.

PyTorch implementation of SFNet This is the implementation of the paper "SFNet: Learning Object-aware Semantic Correspondence". For more information,

CV Lab @ Yonsei University 87 Dec 30, 2022
Saeed Lotfi 28 Dec 12, 2022