-
Notifications
You must be signed in to change notification settings - Fork 45.7k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Perspective transformer 3D evaluation script broken #2239
Comments
Looks like the API of tf.metrics changes in v1.3. Related issue: DrSleep/tensorflow-deeplab-resnet#107 |
This is weird, because you guys said that ptn requieres tf 1.3, right? Anyway, how can we fix the bug? |
@xcyan @arkanath can you reproduce this bug? I was checking but can't find no API changes for tf.metrics.mean_iou |
I found the bug, created a PR to fix it |
@mees The problem is due to this change : line 172 |
I didn't see any related changes in line 172. What do you mean? |
Sure, thanks @xcyan! |
This commit add the assert:
I don't think the assert is nice. |
System information
What is the top-level directory of the model you are using: Ptn directory
Have I written custom code (as opposed to using a stock example script provided in TensorFlow): Yes, I have fixed the first flag issue mees@48ac9b0
OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Ubuntu 14.04
TensorFlow installed from (source or binary): Source
TensorFlow version (use command below): ('v1.3.0-rc1-27-g2784b1c', '1.3.0-rc2')
Bazel version (if compiling from source): 0.5.2
CUDA/cuDNN version: 8.0/5.1
GPU model and memory: Titan X
Exact command to reproduce: CUDA_VISIBLE_DEVICES=0 bazel-bin/eval_ptn '--checkpoint_dir=/home/meeso/models/ptn/my_models/no_finetune' '--inp_dir=/tmp/shapenet_tf/' '--model_name=ptn_finetune'
Describe the problem
Once I train a ptn network, I run the evaluation script (once fixed the first problem with the wrong flags #2236) to compute the IOU. I get an error that the predictions are out of bound. The offending line seems to be here https://github.com/tensorflow/models/blob/master/ptn/metrics.py#L103 but I am not sure where this mapping comes from (3*times predictions -2) and if its correct.
Source code / logs
The text was updated successfully, but these errors were encountered: