For VPU - we use grouped-convolution, but we refrain from using Squeeze-and-excitement (SE) blocks - specifically this includes the following models: EfficientNet-lite / MixNet [76] / GhostNet [21] / MobileNetV3

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )

http://pjreddie.com/darknet/

Other

21.57k stars 7.95k forks source link

For VPU - we use grouped-convolution, but we refrain from using Squeeze-and-excitement (SE) blocks - specifically this includes the following models: EfficientNet-lite / MixNet [76] / GhostNet [21] / MobileNetV3 #5565

Open zyxcambridge opened 4 years ago

zyxcambridge commented 4 years ago

yolov4 use group-conv ?

AlexeyAB commented 4 years ago

No. yolov4 is for GPU. We are exploring different models for yolov4-vpu,, but it seems we will not use: Grouped/Depthwise-conv, SE-blocks, Swish/Mish-activations.