using the ResNet 101 and the pyramid pooling way to detect the road on KITTI raod dataset. It have achived the results as follow:
| Benchmark | MaxF | AP | PRE | REC | FPR | FNR |
|---|---|---|---|---|---|---|
| UM_ROAD | 96.04% | 89.77% | 95.61% | 96.48% | 2.02% | 3.52% |
| UMM_ROAD | 97.03% | 92.36% | 96.36% | 97.70% | 4.06% | 2.30% |
| UU_ROAD | 95.47% | 88.74% | 95.16 % | 95.77% | 1.59% | 4.23% |
This is the link: http://www.cvlibs.net/datasets/kitti/eval_road.php
The project name is RPP. It have the rank of 1st on um_road, 4th on umm_road and 1st on uu_road on the data of Dec 25, 2016.
The repository includes model file, deploy.prototxt, and some pics. You can download and run on justify the model result.
Model download link: https://pan.baidu.com/s/1c4fdixU The caffe should be installed: https://github.com/xllau/PSPNet