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
GeoMol: Torsional Geometric Generation of Molecular 3D Conformer Ensembles

GeoMol: Torsional Geometric Generation of Molecular 3D Conformer Ensembles This repository contains a method to generate 3D conformer ensembles direct

127 Dec 20, 2022
The Noise Contrastive Estimation for softmax output written in Pytorch

An NCE implementation in pytorch About NCE Noise Contrastive Estimation (NCE) is an approximation method that is used to work around the huge computat

Kaiyu Shi 287 Nov 25, 2022
TensorFlow Implementation of Unsupervised Cross-Domain Image Generation

Domain Transfer Network (DTN) TensorFlow implementation of Unsupervised Cross-Domain Image Generation. Requirements Python 2.7 TensorFlow 0.12 Pickle

Yunjey Choi 864 Dec 30, 2022
Exploring Classification Equilibrium in Long-Tailed Object Detection, ICCV2021

Exploring Classification Equilibrium in Long-Tailed Object Detection (LOCE, ICCV 2021) Paper Introduction The conventional detectors tend to make imba

52 Nov 21, 2022
Unofficial implementation of Google's FNet: Mixing Tokens with Fourier Transforms

FNet: Mixing Tokens with Fourier Transforms Pytorch implementation of Fnet : Mixing Tokens with Fourier Transforms. Citation: @misc{leethorp2021fnet,

Rishikesh (ऋषिकेश) 218 Jan 05, 2023
Repository for tackling Kaggle Ultrasound Nerve Segmentation challenge using Torchnet.

Ultrasound Nerve Segmentation Challenge using Torchnet This repository acts as a starting point for someone who wants to start with the kaggle ultraso

Qure.ai 46 Jul 18, 2022
AEI: Actors-Environment Interaction with Adaptive Attention for Temporal Action Proposals Generation

AEI: Actors-Environment Interaction with Adaptive Attention for Temporal Action Proposals Generation A pytorch-version implementation codes of paper:

11 Dec 13, 2022
Implementation of a Transformer, but completely in Triton

Transformer in Triton (wip) Implementation of a Transformer, but completely in Triton. I'm completely new to lower-level neural net code, so this repo

Phil Wang 152 Dec 22, 2022
Discovering Dynamic Salient Regions with Spatio-Temporal Graph Neural Networks

Discovering Dynamic Salient Regions with Spatio-Temporal Graph Neural Networks This is the official code for DyReg model inroduced in Discovering Dyna

Bitdefender Machine Learning 11 Nov 08, 2022
Codes for paper "Towards Diverse Paragraph Captioning for Untrimmed Videos". CVPR 2021

Towards Diverse Paragraph Captioning for Untrimmed Videos This repository contains PyTorch implementation of our paper Towards Diverse Paragraph Capti

Yuqing Song 61 Oct 11, 2022
An API-first distributed deployment system of deep learning models using timeseries data to analyze and predict systems behaviour

Gordo Building thousands of models with timeseries data to monitor systems. Table of content About Examples Install Uninstall Developer manual How to

Equinor 26 Dec 27, 2022
GLaRA: Graph-based Labeling Rule Augmentation for Weakly Supervised Named Entity Recognition

GLaRA: Graph-based Labeling Rule Augmentation for Weakly Supervised Named Entity Recognition

Xinyan Zhao 29 Dec 26, 2022
MG-GCN: Scalable Multi-GPU GCN Training Framework

MG-GCN MG-GCN: multi-GPU GCN training framework. For more information, please read our paper. After cloning our repository, run git submodule update -

Translational Data Analytics (TDA) Lab @GaTech 6 Oct 24, 2022
The official implementation of the CVPR2021 paper: Decoupled Dynamic Filter Networks

Decoupled Dynamic Filter Networks This repo is the official implementation of CVPR2021 paper: "Decoupled Dynamic Filter Networks". Introduction DDF is

F.S.Fire 180 Dec 30, 2022
Implementation of the SUMO (Slim U-Net trained on MODA) model

SUMO - Slim U-Net trained on MODA Implementation of the SUMO (Slim U-Net trained on MODA) model as described in: TODO: add reference to paper once ava

6 Nov 19, 2022
SANet: A Slice-Aware Network for Pulmonary Nodule Detection

SANet: A Slice-Aware Network for Pulmonary Nodule Detection This paper (SANet) has been accepted and early accessed in IEEE TPAMI 2021. This code and

Jie Mei 39 Dec 17, 2022
Kaggle competition: Springleaf Marketing Response

PruebaEnel Prueba Kaggle-Springleaf-master Prueba Kaggle-Springleaf Kaggle competition: Springleaf Marketing Response Competencia de Kaggle: Marketing

1 Feb 09, 2022
Banglore House Prediction Using Flask Server (Python)

Banglore House Prediction Using Flask Server (Python) 🌐 Links 🌐 📂 Repo In this repository, I've implemented a Machine Learning-based Bangalore Hous

Dhyan Shah 1 Jan 24, 2022
Practical tutorials and labs for TensorFlow used by Nvidia, FFN, CNN, RNN, Kaggle, AE

TensorFlow Tutorial - used by Nvidia Learn TensorFlow from scratch by examples and visualizations with interactive jupyter notebooks. Learn to compete

Alexander R Johansen 1.9k Dec 19, 2022