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
A high-performance distributed deep learning system targeting large-scale and automated distributed training.

HETU Documentation | Examples Hetu is a high-performance distributed deep learning system targeting trillions of parameters DL model training, develop

DAIR Lab 150 Dec 21, 2022
Official Code for VideoLT: Large-scale Long-tailed Video Recognition (ICCV 2021)

Pytorch Code for VideoLT [Website][Paper] Updates [10/29/2021] Features uploaded to Google Drive, for access please send us an e-mail: zhangxing18 at

Skye 26 Sep 18, 2022
A simple python stock Predictor

Python Stock Predictor A simple python stock Predictor Demo Run Locally Clone the project git clone https://github.com/yashraj-n/stock-price-predict

Yashraj narke 5 Nov 29, 2021
Simple renderer for use with MuJoCo (>=2.1.2) Python Bindings.

Viewer for MuJoCo in Python Interactive renderer to use with the official Python bindings for MuJoCo. Starting with version 2.1.2, MuJoCo comes with n

Rohan P. Singh 62 Dec 30, 2022
Contains code for Deep Kernelized Dense Geometric Matching

DKM - Deep Kernelized Dense Geometric Matching Contains code for Deep Kernelized Dense Geometric Matching We provide pretrained models and code for ev

Johan Edstedt 83 Dec 23, 2022
Open source repository for the code accompanying the paper 'Non-Rigid Neural Radiance Fields Reconstruction and Novel View Synthesis of a Deforming Scene from Monocular Video'.

Non-Rigid Neural Radiance Fields This is the official repository for the project "Non-Rigid Neural Radiance Fields: Reconstruction and Novel View Synt

Facebook Research 296 Dec 29, 2022
Direct design of biquad filter cascades with deep learning by sampling random polynomials.

IIRNet Direct design of biquad filter cascades with deep learning by sampling random polynomials. Usage git clone https://github.com/csteinmetz1/IIRNe

Christian J. Steinmetz 55 Nov 02, 2022
Pytorch implementation of FlowNet by Dosovitskiy et al.

FlowNetPytorch Pytorch implementation of FlowNet by Dosovitskiy et al. This repository is a torch implementation of FlowNet, by Alexey Dosovitskiy et

Clément Pinard 762 Jan 02, 2023
A minimal yet resourceful implementation of diffusion models (along with pretrained models + synthetic images for nine datasets)

A minimal yet resourceful implementation of diffusion models (along with pretrained models + synthetic images for nine datasets)

Vikash Sehwag 65 Dec 19, 2022
Reference PyTorch implementation of "End-to-end optimized image compression with competition of prior distributions"

PyTorch reference implementation of "End-to-end optimized image compression with competition of prior distributions" by Benoit Brummer and Christophe

Benoit Brummer 6 Jun 16, 2022
EMNLP'2021: Simple Entity-centric Questions Challenge Dense Retrievers

EntityQuestions This repository contains the EntityQuestions dataset as well as code to evaluate retrieval results from the the paper Simple Entity-ce

Princeton Natural Language Processing 119 Sep 28, 2022
Measuring Coding Challenge Competence With APPS

Measuring Coding Challenge Competence With APPS This is the repository for Measuring Coding Challenge Competence With APPS by Dan Hendrycks*, Steven B

Dan Hendrycks 218 Dec 27, 2022
A GPT, made only of MLPs, in Jax

MLP GPT - Jax (wip) A GPT, made only of MLPs, in Jax. The specific MLP to be used are gMLPs with the Spatial Gating Units. Working Pytorch implementat

Phil Wang 53 Sep 27, 2022
Securetar - A streaming wrapper around python tarfile and allow secure handling files and support encryption

Secure Tar Secure Tarfile library It's a streaming wrapper around python tarfile

Pascal Vizeli 2 Dec 09, 2022
TCube generates rich and fluent narratives that describes the characteristics, trends, and anomalies of any time-series data (domain-agnostic) using the transfer learning capabilities of PLMs.

TCube: Domain-Agnostic Neural Time series Narration This repository contains the code for the paper: "TCube: Domain-Agnostic Neural Time series Narrat

Mandar Sharma 7 Oct 31, 2021
Code for "Long Range Probabilistic Forecasting in Time-Series using High Order Statistics"

Long Range Probabilistic Forecasting in Time-Series using High Order Statistics This is the code produced as part of the paper Long Range Probabilisti

16 Dec 06, 2022
🎯 A comprehensive gradient-free optimization framework written in Python

Solid is a Python framework for gradient-free optimization. It contains basic versions of many of the most common optimization algorithms that do not

Devin Soni 565 Dec 26, 2022
[CVPR 2021] MiVOS - Mask Propagation module. Reproduced STM (and better) with training code :star2:. Semi-supervised video object segmentation evaluation.

MiVOS (CVPR 2021) - Mask Propagation Ho Kei Cheng, Yu-Wing Tai, Chi-Keung Tang [arXiv] [Paper PDF] [Project Page] [Papers with Code] This repo impleme

Rex Cheng 106 Jan 03, 2023
Detail-Preserving Transformer for Light Field Image Super-Resolution

DPT Official Pytorch implementation of the paper "Detail-Preserving Transformer for Light Field Image Super-Resolution" accepted by AAAI 2022 . Update

50 Jan 01, 2023
This repository contains codes of ICCV2021 paper: SO-Pose: Exploiting Self-Occlusion for Direct 6D Pose Estimation

SO-Pose This repository contains codes of ICCV2021 paper: SO-Pose: Exploiting Self-Occlusion for Direct 6D Pose Estimation This paper is basically an

shangbuhuan 52 Nov 25, 2022