ID	rn50_in1k_color_goyal
Layers	50
Classes	1000
Width Multiplier	1

ID	rn50_in22k_color_goyal
Layers	50
Classes	22000
Width Multiplier	1

ID	rn50_yfcc100m_color_goyal
Layers	50
Width Multiplier	1

Colorization

facebookresearch / vissl

Last updated on Feb 27, 2021

Colorization ResNet-50 (Goyal19, ImageNet-1K) achieves 82.3% Top 1 Accuracy on ImageNet

Parameters 26 Million

FLOPs 4 Billion

File Size 177.97 MB

Training Data ImageNet

Training Resources 8 NVIDIA V100 GPUs

Training Time

Training Techniques	Colorization, Weight Decay, SGD with Momentum
Architecture	1x1 Convolution, Bottleneck Residual Block, Batch Normalization, Convolution, Global Average Pooling, Residual Block, Residual Connection, ReLU, Max Pooling, Softmax
ID	rn50_in1k_color_goyal
Layers	50
Classes	1000
Width Multiplier	1
SHOW MORE
SHOW LESS

Colorization ResNet-50 (Goyal19, ImageNet-22K) achieves 82.5% Top 1 Accuracy on ImageNet

Parameters 26 Million

FLOPs 4 Billion

File Size 177.97 MB

Training Data ImageNet

Training Resources 8 NVIDIA V100 GPUs

Training Time

Training Techniques	Colorization, Weight Decay, SGD with Momentum
Architecture	1x1 Convolution, Bottleneck Residual Block, Batch Normalization, Convolution, Global Average Pooling, Residual Block, Residual Connection, ReLU, Max Pooling, Softmax
ID	rn50_in22k_color_goyal
Layers	50
Classes	22000
Width Multiplier	1
SHOW MORE
SHOW LESS

Colorization ResNet-50 (Goyal19, YFCC100M) achieves 82.8% Top 1 Accuracy on ImageNet

Parameters 26 Million

FLOPs 4 Billion

File Size 177.97 MB

Training Data YFCC100M

Training Resources 8 NVIDIA V100 GPUs

Training Time

Training Techniques	Colorization, Weight Decay, SGD with Momentum
Architecture	1x1 Convolution, Bottleneck Residual Block, Batch Normalization, Convolution, Global Average Pooling, Residual Block, Residual Connection, ReLU, Max Pooling, Softmax
ID	rn50_yfcc100m_color_goyal
Layers	50
Width Multiplier	1
SHOW MORE
SHOW LESS

README.md

Summary

Colorization is a self-supervision approach that relies on colorization as the pretext task in order to learn image representations. This particular set of models includes improved models for Colorization that employ:

Scaling pre-training data: scaling to 100× more data (YFCC-100M).
Scaling model capacity: scaling up to a higher capacity model, ResNet-50, that shows larger improvements as the data size increases.
Scaling problem complexity: scaling the ‘hardness’; observing higher capacity models show a larger improvement on ‘harder’ tasks.

How do I train this model?

Get started with VISSL by trying one of the Colab tutorial notebooks.

Citation

@article{DBLP:journals/corr/abs-1905-01235,
  author    = {Priya Goyal and
               Dhruv Mahajan and
               Abhinav Gupta and
               Ishan Misra},
  title     = {Scaling and Benchmarking Self-Supervised Visual Representation Learning},
  journal   = {CoRR},
  volume    = {abs/1905.01235},
  year      = {2019},
  url       = {http://arxiv.org/abs/1905.01235},
  archivePrefix = {arXiv},
  eprint    = {1905.01235},
  timestamp = {Mon, 28 Sep 2020 08:19:37 +0200},
  biburl    = {https://dblp.org/rec/journals/corr/abs-1905-01235.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}

@article{DBLP:journals/corr/abs-1803-07728,
  author    = {Spyros Gidaris and
               Praveer Singh and
               Nikos Komodakis},
  title     = {Unsupervised Representation Learning by Predicting Image Rotations},
  journal   = {CoRR},
  volume    = {abs/1803.07728},
  year      = {2018},
  url       = {http://arxiv.org/abs/1803.07728},
  archivePrefix = {arXiv},
  eprint    = {1803.07728},
  timestamp = {Mon, 13 Aug 2018 16:46:04 +0200},
  biburl    = {https://dblp.org/rec/journals/corr/abs-1803-07728.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}

Results

Image Classification on ImageNet

Image Classification

BENCHMARK	MODEL	METRIC NAME	METRIC VALUE	GLOBAL RANK
ImageNet	Colorization ResNet-50 (Goyal19, ImageNet-22K)	Top 1 Accuracy	49.24%	# 313
ImageNet	Colorization ResNet-50 (Goyal19, YFCC100M)	Top 1 Accuracy	47.46%	# 317
ImageNet	Colorization ResNet-50 (Goyal19, ImageNet-1K)	Top 1 Accuracy	40.11%	# 321