lmb-freiburg / deeptam Goto Github PK

Hi, authors,

When will the mapping code be released?

THX!

Hey!
I would like to retrain the model from scratch on my own dataset. Is there a way, or a script to do so?

Thank you for your guidance.

Regards.

Dear authors,

In rgbd_sequence, when tpose is not matched, why R, t are set to identity matrix?

Thanks

Hi,

I tried running the tracking code (deeptam/tracking/examples/example_advanced_sequence.py) but the translational RMSE printed at the end is 679702 m/s. The visualization shows no current image. The weights are loading fine. Any idea what might be going wrong? Here's the output for one of the intermediate frames. Also, every time I run it, there is a slight variation in the RMSE values. Shouldn't it remain the same since we are loading the pre trained weights. Thanks for the help in advance!

when I compile lmbspecialops, there is always an error occurs
decode_flo_op.cc:65:15: error: ‘const class tensorflow::StringPiece’ has no member named ‘starts_with’ if (!data.starts_with("PIEH")) {
I try to trace back, and found that there is a 'starts_with' function in namespace protobuf/google but not in namespace tensorflow.
how to fix it?

Hi,

Thank you for releasing your code. I was able to reproduce the results shown on your paper.

For the training of the tracking part, you mentioned in your paper that

We train on image pairs from the SUN3D dataset and the SUNCG dataset. For SUN3D we sample image pairs with a baseline of up to 40cm. For SUNCG we generate images with normally distributed baselines with standard deviation 15cm and rotation angles with standard deviation 0.15 radians. When sampling an image pair we reject samples with an image overlap of less than 50%. For keyframe depth maps DK, we use the ground truth depth from the datasets during training.

Can you please share the exact data on which you trained your tracking model? or maybe a script to download the data in the required format to train it.

Similarly for the mapping module.

Thank you so much in advance!

How can I extract the depth metric of the keyframe and what will be the units of the depth metric?

Hi Dear authors,

Thanks for publishing your great work, could you give me some pointers to build Dockerfile for the environment? Right now, I can only see TF is needed as the dependency. Does it also require g2o and pavalion like other slam projects? Thank you very much!

example_advanced_sequence.py gives large testing error on the sequence freiburg1_desk2.

May I ask how to get the results in the paper? It seems that the released code is not complete especially for some corner cases.

Hi,

I'm failing on running the make command after cloning.
This is the error I get:

[ 34%] Building NVCC (Device) object lib/CMakeFiles/lmbspecialops.dir/lmbspecialops_generated_warp2d_cuda.cu.o
/home/user/src/DeepTAM/lmbspecialops/src/warp2d_cuda.cu:20:10: fatal error: tensorflow/core/framework/op_kernel.h: No such file or directory
 #include "tensorflow/core/framework/op_kernel.h"
          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
compilation terminated.
CMake Error at lmbspecialops_generated_warp2d_cuda.cu.o.Release.cmake:219 (message):
  Error generating
  /home/user/src/DeepTAM/lmbspecialops/build/lib/CMakeFiles/lmbspecialops.dir//./lmbspecialops_generated_warp2d_cuda.cu.o


lib/CMakeFiles/lmbspecialops.dir/build.make:119: recipe for target 'lib/CMakeFiles/lmbspecialops.dir/lmbspecialops_generated_warp2d_cuda.cu.o' failed
make[2]: *** [lib/CMakeFiles/lmbspecialops.dir/lmbspecialops_generated_warp2d_cuda.cu.o] Error 1
CMakeFiles/Makefile2:90: recipe for target 'lib/CMakeFiles/lmbspecialops.dir/all' failed
make[1]: *** [lib/CMakeFiles/lmbspecialops.dir/all] Error 2
Makefile:94: recipe for target 'all' failed
make: *** [all] Error 2

I have Cuda 10.0 installed because Cuda 8 didn't want to install under Ubuntu 18.04. Also because of that I use Tensorflow 1.14.0

Hi there,

Are you going to release the training code? That would be really helpful.

Thank you very much,
Tassos

Currently, tracking and mapping are separated. I wonder if you can also provide the code for simultaneous tracking and mapping.

Hi,

I'm trying to get DeepTAM running but at the moment I'm stuck with the following error message

KeyError: "Registering two gradient with name 'ScaleInvariantGradient' !(Previous registration was in init /home/adas/DeepTAM/deeptam/tracking/python/deeptam_tracker/tracker.py:231)"

Every other problem was more or less easy to find out but this is very hard and I can't get it worked out.

Here would be the full stack trace if it helps:
https://pastebin.com/2qu72vnJ

Using /home/adas/DeepTAM/lmbspecialops/build/lib/lmbspecialops.so
Traceback (most recent call last):
  File "example_basic.py", line 102, in <module>
    main()
  File "example_basic.py", line 72, in main
    tracker_core = TrackerCore(tracking_module_path,checkpoint,intrinsics)
  File "/home/adas/DeepTAM/deeptam/tracking/python/deeptam_tracker/tracker.py", line 231, in __init__
    self._tracking_mod = load_myNetworks_module_noname(self._tracking_module)
  File "/home/adas/DeepTAM/deeptam/tracking/python/deeptam_tracker/utils/helpers.py", line 44, in load_myNetworks_module_noname
    return load_myNetworks_module(module_name, path_to_myNetworks)
  File "/home/adas/DeepTAM/deeptam/tracking/python/deeptam_tracker/utils/helpers.py", line 32, in load_myNetworks_module
    return importlib.import_module('.'+myNetworks_name, module_name)
  File "/home/adas/.virtualenvs/DeepTAM/lib/python3.6/importlib/__init__.py", line 126, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
  File "<frozen importlib._bootstrap>", line 994, in _gcd_import
  File "<frozen importlib._bootstrap>", line 971, in _find_and_load
  File "<frozen importlib._bootstrap>", line 955, in _find_and_load_unlocked
  File "<frozen importlib._bootstrap>", line 665, in _load_unlocked
  File "<frozen importlib._bootstrap_external>", line 678, in exec_module
  File "<frozen importlib._bootstrap>", line 219, in _call_with_frames_removed
  File "../python/deeptam_tracker/models/networks.py", line 6, in <module>
    from .blocks import *
  File "../python/deeptam_tracker/models/blocks.py", line 1, in <module>
    from .helpers import *
  File "../python/deeptam_tracker/models/helpers.py", line 4, in <module>
    spec.loader.exec_module(foo)
  File "/home/adas/DeepTAM/lmbspecialops/python/lmbspecialops.py", line 356, in <module>
    @ops.RegisterGradient("ScaleInvariantGradient")
  File "/home/adas/.virtualenvs/DeepTAM/lib/python3.6/site-packages/tensorflow/python/framework/ops.py", line 2085, in __call__
    _gradient_registry.register(f, self._op_type)
  File "/home/adas/.virtualenvs/DeepTAM/lib/python3.6/site-packages/tensorflow/python/framework/registry.py", line 62, in register
    (self._name, name, function_name, filename, line_number))
KeyError: "Registering two gradient with name 'ScaleInvariantGradient' !(Previous registration was in __init__ /home/adas/DeepTAM/deeptam/tracking/python/deeptam_tracker/tracker.py:231)"
Tensorflow graph reseted

This happens when running the example_basic.py or example_advanced_sequence.py

Hope you have an idea how to fix it