when I use the pre-trained pose model to predict, the output is same whatever input are

[KerunLi]

when I use the pre-trained pose model to predict, the output is same whatever inputs are.

I use seq 9 and seq 11 to run test_kitti_pose.py, but the output pose.txt are all the same, that is, 000000.txt and 000001.txt has the same content.

I print the input image data, it's not same, but the output pose is same. I normalized the input image data but it doesn't work.

（I use tensorflow-gpu 2.0.0b, cuda 10.0 and python 3.6, and use "import tensorflow.compat.v1 as tf" and "tf.disable_v2_behavior()" replace import tensorflow as tf

thanks a lot for any help.

[Yifan-github]

i have a question.where is the kitti_depth_model.tar?

i can't download the pre-trained model by running the "bash ./models/download_depth_model.sh".

Hope to get a help,thanks a lot.

[KerunLi]

@Yifan-github You can try run wget http://people.eecs.berkeley.edu/~tinghuiz/projects/SfMLearner/models/kitti_depth_model.tar in terminal, then unpack it.
if you still cann't download, maybe it's the problem of network or vpn

[Yifan-github]

i met a ERROR 404: Not Found.Is the file sharing cancelled?If possible, may you share me a copy by email?Thanks so much!My E-mail:820407324@qq.com