MaoYuxin

请问为什么要对batch做循环

2

您好，非常感谢您开源代码。在代码的模型部分，我看到主模型中的forward对每一个batch做了循环，每一次循环独立处理，在我理解中，batch本身就是独立的，那么是否可以不做循环而直接批次处理呢？

Problem while running 'roslaunch bebop_tools joy_teleop.launch'

1

process[bebop/robot_state_publisher-3]: started with pid [25681] [ INFO] [1538823198.803695714]: Initializing nodelet with 4 worker threads. [ INFO] [1538823198.824396477]: [BebopSDK] 06:53:18:824 | Bebop:225 - Bebop Cnstr() [ INFO] [1538823198.824488178]: Nodelet Cstr [...

question about optical flow resize

1

hello antonilo. Thank you for such an outstanding job! At line 89 in file adversarial_learner.py, you performed a resize operation on the optical flow, through tf.image.resize. However, whether the resize...

Code for Correlation Range of 128 pixels.

1

Hello, thank you very much for such a valuable work, after two years, it is still so excellent. In the table2 of the paper, when the Correlation Range is 128px,...

训练模型无法复现结果

7

您好，感谢您开源代码，代码写的很清晰，readme的可操作性也很强。但是我在尝试训练IFRNet_S的时候，发现我训练后的模型无法正确复现论文中的结果。我用了4张3090，完全按照readme中的命令进行的操作。最终只能得到34.45的PSNR。我的train.log如下，想问问可能出现的问题会在哪里，期待您的回复。 [train.log](https://github.com/ltkong218/IFRNet/files/8975261/train.log)

Questions about how to visualize optical flow

2

Hello, thank you for such valuable work. I read and debugged your code, and I have a question about how to visualize the optical flow calculated by the motionsqueeze module....

When running CIFAR10 demo, NAN appears in loss.

14

Hello, I am running the CIFAR10 demo you provided. The script used is as follows. I found that the loss cannot be optimized and NAN appears. Is there any solution...

Training issues

16

Hi, I found a file "landscape_linear1000_16x64x64_shiftT_window148_lr1e-4_ema_100000.pt" in the training script. Is this file the landscape.pt in the open-source model? I am looking forward to your answer, thank you very much.

How many epochs are needed to reproduce the results?

1

Hello, I would like to ask how many epochs are needed to train the imagenet data set to obtain similar results to those in the paper? Looking forward to your...