You Only Look One-level Feature

17 Mar 2021 Qiang Chen Yingming Wang Tong Yang Xiangyu Zhang Jian Cheng Jian Sun

This paper revisits feature pyramids networks (FPN) for one-stage detectors and points out that the success of FPN is due to its divide-and-conquer solution to the optimization problem in object detection rather than multi-scale feature fusion. From the perspective of optimization, we introduce an alternative way to address the problem instead of adopting the complex feature pyramids - {\em utilizing only one-level feature for detection}... (read more)

PDF Abstract

Datasets


Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT BENCHMARK
Object Detection COCO test-dev YOLOF-DC5 box AP 44.3 # 82
AP50 62.9 # 86
AP75 47.5 # 79
APS 24.0 # 96
APM 48.5 # 63
APL 60.4 # 44

Methods used in the Paper


METHOD TYPE
Batch Normalization
Normalization
Max Pooling
Pooling Operations
Average Pooling
Pooling Operations
Sigmoid Activation
Activation Functions
Global Average Pooling
Pooling Operations
Tanh Activation
Activation Functions
Darknet-53
Convolutional Neural Networks
ReLU
Activation Functions
k-Means Clustering
Clustering
Logistic Regression
Generalized Linear Models
Layer Normalization
Normalization
Softplus
Activation Functions
Mish
Activation Functions
Bottom-up Path Augmentation
Feature Extractors
Adam
Stochastic Optimization
Spatial Attention Module
Image Model Blocks
YOLOv3
Object Detection Models
Scaled Dot-Product Attention
Attention Mechanisms
CSPDarknet53
Convolutional Neural Networks
BPE
Subword Segmentation
Residual Connection
Skip Connections
Label Smoothing
Regularization
DropBlock
Regularization
1x1 Convolution
Convolutions
Convolution
Convolutions
CutMix
Image Data Augmentation
Dropout
Regularization
Multi-Head Attention
Attention Modules
Dense Connections
Feedforward Networks
Feedforward Network
Feedforward Networks
Softmax
Output Functions
Spatial Pyramid Pooling
Pooling Operations
Cosine Annealing
Learning Rate Schedules
PAFPN
Feature Extractors
Transformer
Transformers
Focal Loss
Loss Functions
YOLOv4
Object Detection Models
Detr
Object Detection Models
FPN
Feature Extractors
RetinaNet
Object Detection Models