Image Classification (CNN)

image classification and convolutional neural networks

Survey Papers / Repos

• Recent Advances in Convolutional Neural Networks [1512.07108]

• weiaicunzai/awesome-image-classification

Resources

• CS231n: Convolutional Neural Networks for Visual Recognition

• torchvision: official pytorch toolkit for vision

• Cadene/pretrained-models.pytorch: lots of pretrained models in pytorch

• albumentations-team/albumentations: lots of augmentation implementations

• aleju/imgaug: lots of augmentation implementations

Models

Convolutions ([vdumoulin/conv_arithmetic] [1603.07285])

• Vanilla Convolution

• 1x1 Convolution [1312.4400]

• Transpose Convolution [1411.4038]

• Dilated Convolution [1511.07122]

• Deformable Convolution [1703.06211]

Architectures

• LeNet

• AlexNet

• VGG [1409.1556]

• GoogLeNet (Inception v1) [1409.4842]

• Inception v2, Inception v3 [1512.005677]

• ResNet [1512.03385]

• Inception v4, Inception ResNet [1602.07261]

• SqueezeNet [1602.07360]

• WideResNet [1605.07146]

• DenseNet [1608.06993]

• ResNeXt [1611.05431]

• MobileNet [1704.04861]

• NASNet [1707.07012]

• SE-ResNet [1709.01507]

• Res2Net [1904.01169]

• EfficientNet [1905.11946]

• HRNet [1908.07919]

Loss

• Cross Entropy Loss

• Focal Loss

Augmentation

• Conventional: Random Crop, ...

• Cutout [1708.04552]

• AutoAugment [1805.09501]

• Fast AutoAugment [1905.00397]

• RandAugment [1909.13719]

Others

Dataset

• MNIST

• Fashion-MNIST

• CIFAR

• ImageNet

• Open Images Dataset

Tricks

• Mixup [1710.09412]

• Pseudo Labeling

• Data Distillation [1712.04440]

Task

• Vanilla Image Classification

• Few-shot Image Classification

• Zero-shot Image Classification