Pyramidal Convolution: Rethinking Convolutional Neural Networks for Visual Recognition

20 Jun 2020Ionut Cosmin DutaLi LiuFan ZhuLing Shao

This work introduces pyramidal convolution (PyConv), which is capable of processing the input at multiple filter scales. PyConv contains a pyramid of kernels, where each level involves different types of filters with varying size and depth, which are able to capture different levels of details in the scene... (read more)

PDF Abstract
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT BENCHMARK
Semantic Segmentation ADE20K PyConvSegNet-152 Validation mIoU 45.99 # 6
Test Score 0.5652 # 1
Semantic Segmentation ADE20K val PyConvSegNet-152 mIoU 45.99 # 6
Pixel Accuracy 82.49 # 1
Image Classification ImageNet PyConvResNet-101 Top 1 Accuracy 81.49% # 45
Top 5 Accuracy 95.72% # 30
Number of params 42.3M # 30

Methods used in the Paper