He Dai

Failed to import deform_conv_cuda

14

Hi, @U @haofeixu , I meet the following trouble when I run your model. ``` Traceback (most recent call last): File "inf_time.py", line 8, in import nets File "/home/lthpc/PRO/aanet-master/nets/__init__.py", line...

Inference time evaluation input size and TX2 backend

1

Hi@v@mileyan, I want to know if the input image size is 1242x375 or 1232x368 since 1242x375 cannot run forward through the model. Another question is the same as #18. Could...

Where is the split.txt

1

Hi, it's nice to see your code released. Excellent work! In the instruction, there is an argument when evaluating kitti2015, which is --split_file checkpoint/kitti2015_ck/split.txt I cannot see the file after...

作者你好，立体视觉（stereo matching）没找到，难道它是自由的吗

3

后面会提供吗

sync_bn

4

作者你好，我正在我的一个立体匹配的工程里用sync_bn。但是在使用syncbn的时候跑总是跑几个iter就卡住（显存仍然占用，但什么反应都没有），我现在的环境是ubuntu+RTX 2080Ti+cuda10，syncbn用法也是参照你的代码，请问是怎么回事 #23 我似乎跟作者你在这里面的情况一样

3-px validation error calculation method

3

Hi! @JiaRenChang In your readme.md, you said Note that the reported 3-px validation errors were calculated using KITTI's official Matlab code, not our code I've checked your 3-px validation error...

Hey, @NoamRosenberg , I found that there were two standard convs in DilConv while separate=False. Why would u do that?

Thanks for your great jobs again! I'm confused about the 'self.ops_index' in 'new_model.py'. I guess it should be fetched in the 'self.cell_arch' (which is a tensor containing branch index and...

Hi, @tlkvstepan , it seems there are only 3PE results of KITTI online benchmark included in your paper. Do you have the full results of KITTI online benchmark or the...