HarDNet: A Low Memory Traffic Network

State-of-the-art neural network architectures such as ResNet, MobileNet, and DenseNet have achieved outstanding accuracy over low MACs and small model size counterparts. However, these metrics might not be accurate for predicting the inference time... (read more)

PDF Abstract ICCV 2019 PDF ICCV 2019 Abstract
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT BENCHMARK
Real-Time Semantic Segmentation Cityscapes test U-HarDNet-70 mIoU 75.9% # 3
Time (ms) 18.8 (1080Ti) # 23
Frame (fps) 53 (1080Ti) # 27
Real-Time Object Detection COCO SSD512-HarDNet85 MAP 35.1 # 18
FPS 39 # 10
inference time (ms) 25 # 7
Real-Time Object Detection COCO CenterNet HarDNet-85 MAP 43.6 # 11
FPS 45 # 7
inference time (ms) 22 # 5

Methods used in the Paper


METHOD TYPE
ReLU
Activation Functions
Batch Normalization
Normalization
Bottleneck Residual Block
Skip Connection Blocks
Residual Connection
Skip Connections
Convolution
Convolutions
Residual Block
Skip Connection Blocks
Average Pooling
Pooling Operations
Concatenated Skip Connection
Skip Connections
ResNet
Convolutional Neural Networks
Global Average Pooling
Pooling Operations
Dense Block
Image Model Blocks
Kaiming Initialization
Initialization
1x1 Convolution
Convolutions
Dropout
Regularization
Dense Connections
Feedforward Networks
Max Pooling
Pooling Operations
Softmax
Output Functions
DenseNet
Convolutional Neural Networks