Compounding the Performance Improvements of Assembled Techniques in a Convolutional Neural Network

17 Jan 2020Jungkyu LeeTaeryun WonTae Kwan LeeHyemin LeeGeonmo GuKiho Hong

Recent studies in image classification have demonstrated a variety of techniques for improving the performance of Convolutional Neural Networks (CNNs). However, attempts to combine existing techniques to create a practical model are still uncommon... (read more)

PDF Abstract
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK USES EXTRA
TRAINING DATA
RESULT LEADERBOARD
Fine-Grained Image Classification FGVC Aircraft Assemble-ResNet-FGVC-50 Accuracy 92.4% # 15
Fine-Grained Image Classification Food-101 Assemble-ResNet-FGVC-50 Accuracy 92.5 # 2
Image Classification ImageNet Assemble-ResNet152 Top 1 Accuracy 84.2% # 23
Fine-Grained Image Classification Oxford 102 Flowers Assemble-ResNet Accuracy 98.9% # 4
Fine-Grained Image Classification Oxford-IIIT Pets Assemble-ResNet-FGVC-50 Top-1 Error Rate 5.7% # 4
Accuracy 94.3% # 5
Fine-Grained Image Classification SOP Assemble-ResNet-FGVC-50 [email protected] 85.9 # 1
Fine-Grained Image Classification Stanford Cars Assemble-ResNet-FGVC-50 Accuracy 94.4% # 12

Methods used in the Paper


METHOD TYPE
Convolution
Convolutions
Residual Connection
Skip Connections
Random Resized Crop
Image Data Augmentation
Random Horizontal Flip
Image Data Augmentation
SGD with Momentum
Stochastic Optimization
Weight Decay
Regularization
Batch Normalization
Normalization
Linear Warmup With Cosine Annealing
Learning Rate Schedules
ResNet-D
Convolutional Neural Networks
1x1 Convolution
Convolutions
Average Pooling
Pooling Operations
Residual Block
Skip Connection Blocks
Bottleneck Residual Block
Skip Connection Blocks
ReLU
Activation Functions
Max Pooling
Pooling Operations
Kaiming Initialization
Initialization
Global Average Pooling
Pooling Operations
Assemble-ResNet
Convolutional Neural Networks
AutoAugment
Image Data Augmentation
Cutout
Image Data Augmentation
ColorJitter
Image Data Augmentation
LSTM
Recurrent Neural Networks
Label Smoothing
Regularization
Mixup
Image Data Augmentation
DropBlock
Regularization
Xavier Initialization
Initialization
Dense Connections
Feedforward Networks
Big-Little Module
Skip Connection Blocks
Selective Kernel
Image Model Blocks
Selective Kernel Convolution
Convolutions
Anti-Alias Downsampling
Downsampling
Softmax
Output Functions
Dilated Convolution
Convolutions
Linear Layer
Feedforward Networks
Channel-wise Soft Attention
Attention Mechanisms
Tanh Activation
Activation Functions
Sigmoid Activation
Activation Functions
MobileNetV1
Convolutional Neural Networks
Depthwise Separable Convolution
Convolutions
Depthwise Convolution
Convolutions
Pointwise Convolution
Convolutions