根据论文写配置

[2805651606]

0%| | 0/160000 [00:00<?, ?it/s]
Traceback (most recent call last):
File "/home/ubuntu/miniconda3/envs/xfeat/lib/python3.8/runpy.py", line 194, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/home/ubuntu/miniconda3/envs/xfeat/lib/python3.8/runpy.py", line 87, in _run_code
exec(code, run_globals)
File "/data3/WSX/accelerated_features/modules/training/train_28.py", line 371, in
trainer.train()
File "/data3/WSX/accelerated_features/modules/training/train_28.py", line 239, in train
p1 = torch.cat([p1s, p1], dim=0)
RuntimeError: Sizes of tensors must match except in dimension 0. Expected size 600 but got size 608 for tensor number 1 in the list.
我看论文里面图片格式是800，600.我改成之后，运行不了。

[Xiaole-An]

HI @2805651606 除了在train里面改动尺寸外，MegaDepthDataset类里也需要改动，你看一下是不是忘记改了。要保证megadepth和warpedCOCO尺寸一样的

[2805651606]

好的谢谢。解决了。
我用你最新发布的评估代码，然后
R_err: 0.07105973478560063, t_err: 0.3281071596782389, Inliers: 0
R_err: 0.13242374205641105, t_err: 0.5692219281803291, Inliers: 0
R_err: 2.0303554516149402, t_err: 9.69551256971788, Inliers: 0
R_err: 1.112681358295564, t_err: 6.819630126296131, Inliers: 0
R_err: 17.15245130998925, t_err: 49.769504454113004, Inliers: 0
R_err: 0.06970964298552414, t_err: 0.2497385048637756, Inliers: 0
R_err: 0.37633511515468315, t_err: 0.4877891889008318, Inliers: 0
R_err: 0.05859537991632689, t_err: 0.28733090161078995, Inliers: 0
R_err: 1.3132266746441483, t_err: 4.702368153727258, Inliers: 0
R_err: 0.0821689290087214, t_err: 0.2963301226338091, Inliers: 0
R_err: 0.11980112699062619, t_err: 0.7924706402811298, Inliers: 0
R_err: 0.3103344789991813, t_err: 1.9342218570046488, Inliers: 0
一个匹配点也找不到，这是因为我之前训练用的图片分辨率用的800*608的缘故？