search

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
http://pjreddie.com/darknet/
Other
21.68k stars 7.96k forks source link

One-dimensional convolution kernel #2067

Open Hello526 opened 5 years ago

Hello526 commented 5 years ago

@AlexeyAB Hi, Do you know how to generate a one-dimensional convolution kernel? In general, the convolution kernels used in YOLOV3 are 3 x 3 and 1 x 1. Can I use the cuDNN library to create a 1×3, 1×5 one-dimensional convolution kernel? I think maybe I can find the answer in https://github.com/AlexeyAB/darknet/blob/master/src/convolutional_layer.c

AlexeyAB commented 5 years ago

@Hello526 Currently you can't do this. I will think to add it.

Hello526 commented 5 years ago

@AlexeyAB Thank you for your kind attention. Is it possible to create a 1×n filter through the cuDNN library?