Simple Baselines for Human Pose Estimation and Tracking

There has been significant progress on pose estimation and increasing interests on pose tracking in recent years. At the same time, the overall algorithm and system complexity increases as well, making the algorithm analysis and comparison more difficult... (read more)

PDF Abstract ECCV 2018 PDF ECCV 2018 Abstract

Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT BENCHMARK
Keypoint Detection COCO ResNet-50 Validation AP 72.2 # 5
Keypoint Detection COCO test-challenge Simple Base+* AR 80.5 # 2
ARM 75.3 # 2
AP 74.5 # 2
AP50 90.9 # 2
AP75 80.8 # 2
APL 87.5 # 2
AR50 95.1 # 2
AR75 86.3 # 2
ARL 82.9 # 2
Keypoint Detection COCO test-dev Simple Base APL 80.0 # 6
APM 70.3 # 5
AP50 91.9 # 5
AP75 81.1 # 5
AR 79.0 # 5
Keypoint Detection COCO test-dev Simple Base+* APL 82.7 # 2
APM 73.0 # 2
AP50 92.4 # 4
AP75 84.0 # 2
AR 81.5 # 3
AR50 95.8 # 2
AR75 88.2 # 1
ARL 87.2 # 1
ARM 77.4 # 2
Pose Estimation COCO test-dev Flow-based (ResNet-152) AP 73.7 # 7
AP50 91.9 # 6
AP75 81.1 # 7
APL 80 # 6
APM 70.3 # 7
AR 79 # 7
Pose Tracking PoseTrack2017 MSRA (FlowTrack) MOTA 57.81 # 4
mAP 74.57 # 2
Pose Tracking PoseTrack2018 MSRA MOTA 61.37 # 2
mAP 74.03 # 1

Methods used in the Paper


METHOD TYPE
Average Pooling
Pooling Operations
Residual Connection
Skip Connections
ReLU
Activation Functions
1x1 Convolution
Convolutions
Batch Normalization
Normalization
Bottleneck Residual Block
Skip Connection Blocks
Global Average Pooling
Pooling Operations
Residual Block
Skip Connection Blocks
Kaiming Initialization
Initialization
Max Pooling
Pooling Operations
Convolution
Convolutions
ResNet
Convolutional Neural Networks