I used the parameters from the file args_res50_kitti_192x640_train to train on the Kitti dataset. The training proceeds normally, but when visualizing the output of the model, I found that the images are completely black when using Resnet_lite as the backbone, and green when using Resnet. Has anyone encountered a similar issue? How was it resolved?
I used the parameters from the file args_res50_kitti_192x640_train to train on the Kitti dataset. The training proceeds normally, but when visualizing the output of the model, I found that the images are completely black when using Resnet_lite as the backbone, and green when using Resnet. Has anyone encountered a similar issue? How was it resolved?