BlitzNet: A Real-Time Deep Network for Scene Understanding @KarolMajek
BlitzNet: A Real-Time Deep Network for Scene Understanding @KarolMajek
Uploaded September 2018 | Updated September 2026, 1 week ago
Code: github.com/dvornikita/blitznet

Input 4K video: bit.ly/4KDDPL_WAW2
BlitzNet: A Real-Time Deep Network for Scene UnderstandingYOLOR ssss DWT 2048x1152Cascade Mask RCNN X101 64x4d FPN 20e (MMdetection)A2D2 DeepLabv3+ R101 - 1000x1000px input PyTorchTensorflow Monodepth - Kitti dataset - Tesla K40cSETR MLA 480x480 80k Pascal Context batch 16Detectron2: Panoptic FPN R101- Panoptic segmentationFaster RCNN X101 64x4d FPN 2xSETR PUP 768x768 80k Cityscapes batch 8Detectron2: Faster RCNN R50 DC5 3x - COCO - Object Detection Tesla V100In-Place Activated BatchNorm - Mapillary Vistas - WideResNet38 + DeepLab3 segmentationFCOS: Fully Convolutional One-Stage Object Detection (ICCV19)
Karol Majek |

BlitzNet: A Real-Time Deep Network for Scene Understanding

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER