基于 FCN 网络对道路的2分割实战【包含数据集、完整代码、训练好的结果、权重文件等等】

[train hyper-parameters: Namespace(backbone='resnet50', base_size=300, batch_size=8, crop_size=256, epochs=5, img_f='.jpg', lr=0.001, lrf=0.01, mask_f='_mask.png', pretrained=True, visTrainSets=True)] [epoch: 1] global correct: 0.9678 precision: ['0.9827', '0.8642'] recall: ['0.9805', '0.8780'] IoU: ['0.9639', '0.7716'] mean IoU: 0.8677 [epoch: 2] global correct: 0.9566 precision: ['0.9683', '0.8630'] recall: ['0.9826', '0.7728'] IoU: ['0.9521', '0.6884'] mean IoU: 0.8202 [epoch: 3] global correct: 0.9692 precision: ['0.9926', '0.8280'] recall: ['0.9721', '0.9489'] IoU: ['0.9651', '0.7927'] mean IoU: 0.8789 [epoch: 4] global correct: 0.9739 precision: ['0.9821', '0.9125'] recall: ['0.9882', '0.8727'] IoU: ['0.9707', '0.8053'] mean IoU: 0.8880 [epoch: 5] global correct: 0.9764 precision: ['0.9859', '0.9087'] recall: ['0.9872', '0.8998'] IoU: ['0.9734', '0.8252'] mean IoU: 0.8993