LEGO
LEGO copied to clipboard
zhenheny
LEGO: Learning Edge with Geometry all at Once by Watching Videos
I have spent a lot of time studying this paper and this code. However, similar to this thread: https://github.com/zhenheny/LEGO/issues/4, I cannot replicate the results of the paper. I am training...
In depth2normal_tf.py, line 121, you do `normal_vector = normalize_l2(normals0)`. Your final normal vector should be some weighted average of the four resulting cross products normalized by l2 norm, right? line...
when I enter cityscapes ,i sign in and found "leftImg8bit_sequence_trainvaltest" dataset. where should I contact to download "this" datasets?The cityscapes page have some direct datasets download link,but some not.
Hi. Thanks for your contribution. These days I am training your network. The total loss fluctuates between 1.5 and 3 after 10 epochs while the edge_loss is fixed at 0.00244~....
Hi, thank you for sharing this work. Your AAAI2018 paper had experiments on NYUv2 (study room) and shown reasonable results. This repo contains code of NYUv2 but your CVPR2018 paper...
`## evaluation for kitti dataset if dataset_name in ["kitti", 'cityscapes']: modes = ["kitti"] root_img_path = "/home/zhenheng/datasets/kitti/" normal_gt_path = "/home/zhenheng/works/unsp_depth_normal/depth2normal/eval/kitti/gt_nyu_fill_depth2nornmal_tf_mask/" normal_gt_path = "/home/zhenheng/datasets/kitti/kitti_normal_gt_monofill_mask/" input_intrinsics = pickle.load( open("/home/zhenheng/datasets/kitti/intrinsic_matrixes.pkl", 'rb'))` That's a piece...
