timostoff / event_cnn_minimal Goto Github PK

View Code? Open in Web Editor NEW

80.0 80.0 24.0 8.62 MB

Minimal code for loading models trained for ECCV'20

Python 97.64% Shell 2.36%

event_cnn_minimal's People

Contributors

Stargazers

Watchers

event_cnn_minimal's Issues

Evaluation code

is evaluation code available with the mentioned pretrained model

link to training data

Congratulations to this great work.

You stated that the training data is open-source, as part of your contribution.

However, when I follow your link for datasets, it seems that it only shows the HQF dataset? Where can I find the training data.

Thanks

Nice work！@TimoStoff
Due to the ubuntu version is 18.04, I got a few questions of converting bag files to h5 files. I don't know if I have the honor to get access to the HQF dataset by HDF5 format directly rather than rosbag format. THKS

Training time

Hello! I have a question about time.
gpu=1, batch_size=2, sequence_length=40
Training on the data set simulated by coco according to the above parameters, how long did it take you to get the best results?

Use cv_bridge and rosbag from conda-forge

Hi Timo,
Just to let you know that these packages are now available from conda-forge and a separate ROS Installation should not be necessary anymore.

Best, Tobi

error when loading the pretrained model

Thank you for your outstanding work!

Problems encountered when training data

Hello!
Thank you for your outstanding work!

I successfully downloaded the HQF dataset and configured the environment you introduced.

But I encountered a problem when cloning the repo and submodules.
fatal: The remote branch inference is not found in the upstream origin
Then I manually downloaded the branch code and put it into the folder.

I want to retrain your network using the HQF dataset to replicate your work. I have converted all the rosbag data to h5 format.
Due to my stupidity, I did not understand what is contained in the data_file /tmp/extracted_data/data_file.txt in the config.json file.
Following the prompt, I filled in the location of the h5 format dataset here, but encountered an error.

~/event_cnn_minimal-master$ python train.py --config config/reconstruction.json 
Traceback (most recent call last):
  File "train.py", line 107, in <module>
    main(config)
  File "train.py", line 56, in main
    data_loader = config.init_obj('data_loader', module_data)
  File "/home/gp/event_cnn_minimal-master/parse_config.py", line 93, in init_obj
    return getattr(module, module_name)(*args, **module_args)
  File "/home/gp/event_cnn_minimal-master/data_loader/data_loaders.py", line 25, in __init__
    dataset = concatenate_datasets(data_file, SequenceDataset, sequence_kwargs)
  File "/home/gp/event_cnn_minimal-master/utils/data.py", line 38, in concatenate_datasets
    data_paths = pd.read_csv(data_file, header=None).values.flatten().tolist()
  File "/home/gp/anaconda3/envs/event_cnn/lib/python3.7/site-packages/pandas/io/parsers.py", line 605, in read_csv
    return _read(filepath_or_buffer, kwds)
  File "/home/gp/anaconda3/envs/event_cnn/lib/python3.7/site-packages/pandas/io/parsers.py", line 457, in _read
    parser = TextFileReader(filepath_or_buffer, **kwds)
  File "/home/gp/anaconda3/envs/event_cnn/lib/python3.7/site-packages/pandas/io/parsers.py", line 814, in __init__
    self._engine = self._make_engine(self.engine)
  File "/home/gp/anaconda3/envs/event_cnn/lib/python3.7/site-packages/pandas/io/parsers.py", line 1045, in _make_engine
    return mapping[engine](self.f, **self.options)  # type: ignore[call-arg]
  File "/home/gp/anaconda3/envs/event_cnn/lib/python3.7/site-packages/pandas/io/parsers.py", line 1893, in __init__
    self._reader = parsers.TextReader(self.handles.handle, **kwds)
  File "pandas/_libs/parsers.pyx", line 518, in pandas._libs.parsers.TextReader.__cinit__
  File "pandas/_libs/parsers.pyx", line 717, in pandas._libs.parsers.TextReader._get_header
  File "pandas/_libs/parsers.pyx", line 814, in pandas._libs.parsers.TextReader._tokenize_rows
  File "pandas/_libs/parsers.pyx", line 1943, in pandas._libs.parsers.raise_parser_error
UnicodeDecodeError: 'utf-8' codec can't decode byte 0x89 in position 0: invalid start byte
== Timing statistics ==

I should have filled in the wrong parameters.

Looking forward to your reply, thank you for your help.

FireNet+

This is more of a question than an issue.

Besides E2VID+, could you also provide your FireNet+ model?

Thanks a lot in advance!

Different evaluation result using the pretrained weight!

I am trying to evaluate the pretrained model reconstruction_model.pth based on the inference.py. But the result is much different to that in paper even when only looking at the MSE? Please help!

Below is my results on HQF:

+--------------------------------------------------------------------------------------------------------------+
|                                                Results on HQF                                                |
+-------------------------+-----------------------------+--------------------------+---------------------------+
|         Sequence        | p_loss/reconstruction_model | mse/reconstruction_model | ssim/reconstruction_model |
+-------------------------+-----------------------------+--------------------------+---------------------------+
|     bike_bay_hdr.h5     |            0.298            |          0.036           |            0.48           |
+-------------------------+-----------------------------+--------------------------+---------------------------+
|         boxes.h5        |            0.264            |          0.049           |           0.497           |
+-------------------------+-----------------------------+--------------------------+---------------------------+
|       desk_fast.h5      |            0.211            |          0.032           |           0.584           |
+-------------------------+-----------------------------+--------------------------+---------------------------+
|         desk.h5         |            0.191            |          0.028           |           0.567           |
+-------------------------+-----------------------------+--------------------------+---------------------------+
|    desk_hand_only.h5    |            0.344            |          0.045           |           0.567           |
+-------------------------+-----------------------------+--------------------------+---------------------------+
|       desk_slow.h5      |            0.227            |          0.038           |            0.6            |
+-------------------------+-----------------------------+--------------------------+---------------------------+
|  engineering_posters.h5 |            0.314            |          0.041           |            0.48           |
+-------------------------+-----------------------------+--------------------------+---------------------------+
|  high_texture_plants.h5 |            0.163            |          0.026           |           0.544           |
+-------------------------+-----------------------------+--------------------------+---------------------------+
|    poster_pillar_1.h5   |            0.269            |          0.032           |           0.467           |
+-------------------------+-----------------------------+--------------------------+---------------------------+
|    poster_pillar_2.h5   |            0.239            |          0.026           |            0.49           |
+-------------------------+-----------------------------+--------------------------+---------------------------+
| reflective_materials.h5 |            0.288            |           0.04           |           0.497           |
+-------------------------+-----------------------------+--------------------------+---------------------------+
|  slow_and_fast_desk.h5  |            0.236            |          0.041           |           0.562           |
+-------------------------+-----------------------------+--------------------------+---------------------------+
|       slow_hand.h5      |            0.317            |          0.048           |           0.489           |
+-------------------------+-----------------------------+--------------------------+---------------------------+
|      still_life.h5      |            0.247            |          0.038           |           0.525           |
+-------------------------+-----------------------------+--------------------------+---------------------------+
|           Mean          |      0.2577142857142857     |   0.03714285714285714    |     0.5249285714285714    |
+-------------------------+-----------------------------+--------------------------+---------------------------+

While the result in paper is:

timostoff / event_cnn_minimal Goto Github PK

event_cnn_minimal's People

Contributors

Stargazers

Watchers

Forkers

event_cnn_minimal's Issues

Evaluation code

link to training data

question about H5 files

Training time

Use cv_bridge and rosbag from conda-forge

error when loading the pretrained model

Thank you for your outstanding work!

Problems encountered when training data

FireNet+

Different evaluation result using the pretrained weight!

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent