Unofficial implementation of the model architecture of Fast-SCNN. Real-time Semantic Segmentation and mobile friendly memory consumption.
Tested with Python 3.6 and Pytorch '1.0.1.post2'
from fast_scnn import Fast_SCNN model = Fast_SCNN(input_channel=3, num_classes=10)
Results with Fast-SCNN
Inference Speed (fps) on Cityscapes
mIoU on Cityscapes
Subscribe to Python Awesome
Get the latest posts delivered right to your inbox