kobiso / computer-vision-leaderboard

Comparison of famous convolutional neural network models

Computer-Vision-Leaderboard

The goal of this repository is:

• To keep on track of state-of-the-art (SoTA) on each vision task and new CNN architectures
• To see the comparison of famous CNN models at a glance (performance, speed, size, etc.)
• To access their research papers and implementations on different frameworks

Leaderboards

Leaderboards are provided in the below link for better accessibility and usability.

• ImageNet Classification Leaderboard
• Stanford Online Products Retrieval Leaderboard
• CARS196 Retrieval Leaderboard
• CUB-200-2011 Retrieval Leaderboard
• In-shop Clothes Retrieval Leaderboard

Updates

ImageNet Classification Leaderboard

• 181022. Update MnasNet and Big-Little Net
• 181024. Create the ImageNet Classification Leaderboard webpage for better accessibility and usability.
• 181126. Update A^2-Nets, ResNet variants from GLUON-CV, and ChannelNets
• 181130. Update FishNet and osmr/imgclsmob repo
• 190326. Update SKNet paper
• 190505. Update RandWire-WS, Res2Net, and MultiGrain papers
• 190524. Update ACNet, iSQRT-COV-Net, MobileNetV3, OctConv, and AAConv papers
• 190610. Update EfficientNet, and GPipe papers

Stanford Online Products Retrieval Leaderboard

• 181222. Create the Stanford Online Products Retrieval Leaderboard
• 190326. Update CGD paper
• 190415. Update HDML and EPSHN paper

CARS196 Retrieval Leaderboard

• 181222. Create the CARS196 Retrieval Leaderboard
• 190326. Update CGD paper
• 190415. Update HDML and EPSHN paper

CUB-200-2011 Retrieval Leaderboard

• 181222. Create the CUB-200-2011 Retrieval Leaderboard
• 190326. Update CGD paper
• 190415. Update HDML and EPSHN paper

In-shop Clothes Retrieval Leaderboard

• 190326. Create the In-shop Clothes Retrieval Leaderboard
• 190326. Update CGD paper
• 190415. Update HDML and EPSHN paper

Related Resources

Check the other good resources about CNN models

• Caffe-model
• TensorNets
• DeepDetect : Open Source Deep Learning Server & API
• Pretrained models for Pytorch
• Pretrained models for Chainer & Gluon
• Netscope CNN Analyzer
• Memory consumption and FLOP count

Author

Byung Soo Ko / [email protected]

computer-vision-leaderboard's Issues

[ImageNet] Exploring the Limits of Weakly Supervised Pretraining

• Paper: https://arxiv.org/abs/1805.00932
• PyTorch: https://github.com/facebookresearch/WSL-Images
  

[ImageNet] Drop an Octave: Reducing Spatial Redundancy in Convolutional Neural Networks with Octave Convolution

• Paper: https://arxiv.org/abs/1904.05049
• MXNet: https://github.com/facebookresearch/OctConv
• PyTorch: https://github.com/ThoroughImages/OctConv

[ImageNet] Deep CNNs Meet Global Covariance Pooling: Better Representation and Generalization

Deep CNNs Meet Global Covariance Pooling: Better Representation and Generalization

• Paper: https://arxiv.org/abs/1904.06836

[ImageNet] FBNet: Hardware-Aware Efficient ConvNet Design via Differentiable Neural Architecture Search

• Paper: https://arxiv.org/abs/1812.03443
• MXNet: https://github.com/JunrQ/NAS/tree/master/fbnet
• PyTorch: https://github.com/JunrQ/NAS/tree/master/fbnet-pytorch

The page is down

I can't see the tables and the the architectures results.

MobileNetV3

Searching for MobileNetV3：https://arxiv.org/abs/1905.02244
github(PyTorch)：https://github.com/xiaolai-sqlai/mobilenetv3

[ImageNet] Spatial Group-wise Enhance: Improving Semantic Feature Learning in Convolutional Networks

• Paper: https://arxiv.org/abs/1905.09646
• PyTorch: https://github.com/implus/PytorchInsight
  

[Retrieval] Divide and Conquer the Embedding Space for Metric Learning

• Paper: https://arxiv.org/abs/1906.05990
• Code: https://github.com/CompVis/metric-learning-divide-and-conquer

[ImageNet] Making Convolutional Networks Shift-Invariant Again

• Paper: https://arxiv.org/abs/1904.11486
• PyTorch: https://github.com/adobe/antialiased-cnns
  

[ImageNet] Res2Net: A New Multi-scale Backbone Architecture

Res2Net: A New Multi-scale Backbone Architecture

• paper: https://arxiv.org/abs/1904.01169

[Retrieval] Metric learning: cross-entropy vs. pairwise losses

• Paper: https://arxiv.org/abs/2003.08983
• Code: https://github.com/jeromerony/dml_cross_entropy

[ImageNet] EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks

• Paper: https://arxiv.org/abs/1905.11946
• TensorFlow: https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet
• MXNet: https://github.com/mnikitin/EfficientNet

[ImageNet] High Frequency Residual Learning for Multi-Scale Image Classification

• Paper: https://arxiv.org/abs/1905.02649
  
  

[ImageNet] Single-Path NAS: Device-Aware Efficient ConvNet Design

• Paper: https://arxiv.org/abs/1905.04159
• Tensorflow: https://github.com/dstamoulis/single-path-nas
  

[ImageNet] ChamNet: Towards Efficient Network Design through Platform-Aware Model Adaptation

• Paper: https://arxiv.org/abs/1812.08934
• PyTorch: https://github.com/facebookresearch/mobile-vision

[Retrieval] The General Pair-based Weighting Loss for Deep Metric Learning

• Paper: https://arxiv.org/abs/1905.12837

Incorrect Efficientnets' FLOPs in ImageNet Classification Leaderboard

Hi, I think in ImageNet Classification Leaderboard, Efficientnet-B7' FLOPS should be 37G (37B in the paper) instead of 37000G. Same for other versions.

[ImageNet] All You Need is a Few Shifts: Designing Efficient Convolutional Neural Networks for Image Classification

• Paper: https://arxiv.org/abs/1903.05285
  
  

[ImageNet] Attention Augmented Convolutional Networks

• Paper: https://arxiv.org/abs/1904.09925
• PyTorch: https://github.com/leaderj1001/Attention-Augmented-Conv2d
• Keras: https://github.com/titu1994/keras-attention-augmented-convs

No pytorch repository for EfficientNet

There is a Pytorch link for the EfficientNet, however it appears that it is linked to a Keras implementation at https://github.com/qubvel/efficientnet

[ImageNet] Exploring Randomly Wired Neural Networks for Image Recognition

Exploring Randomly Wired Neural Networks for Image Recognition

• paper: https://arxiv.org/abs/1904.01569

[ImageNet] Fixing the train-test resolution discrepancy

• Paper : https://arxiv.org/abs/1906.06423
  -Code PyTorch : https://github.com/facebookresearch/FixRes

[Retrieval] Multi-Similarity Loss with General Pair Weighting for Deep Metric Learning

• Paper: https://arxiv.org/abs/1904.06627

[ImageNet] Adaptively Connected Neural Networks

• Paper: https://arxiv.org/abs/1904.03579
• TensorFlow: https://github.com/wanggrun/Adaptively-Connected-Neural-Networks
  

[ImageNet] MultiGrain: a unified image embedding for classes and instances

MultiGrain: a unified image embedding for classes and instances

• paper: https://arxiv.org/abs/1902.05509