caizhongang / waymo_kitti_converter Goto Github PK
View Code? Open in Web Editor NEWA toolkit for Waymo Open Dataset <-> KITTI conversions
A toolkit for Waymo Open Dataset <-> KITTI conversions
Hi,
First of all, thank you very much for making this tool, it has saved me a significant amount of time.
However, I just wanted to inquire, I have some LiDAR detectors that are trained on the KITTI dataset,
but to evaluate an idea, I must use them on Waymo dataset.
since the coordinate frames are different, the bounding boxes are not around the correct objects. Any ideas how I may be able to convert the Waymo reference frame to kitti?
More specifically, I am referring to step 3 here.
https://github.com/open-mmlab/OpenPCDet/blob/master/docs/DEMO.md
I am under the impressions that I must first load the bin files into a np array and then apply some transformation but I am really unsure. I would greatly appreciate any help.
Thank you.
Hi,
When I try pip3 install waymo-open-dataset-tf-2-1-0==1.2.0 --user, I met an error that
ERROR: Could not find a version that satisfies the requirement waymo-open-dataset-tf-2-1-0==1.2.0 (from versions: none)
ERROR: No matching distribution found for waymo-open-dataset-tf-2-1-0==1.2.0
But I have installed the tf2.1 as you recommend. Thank you for help
First of all thanks for the great tool. But I got an error while using small_gt_generator.py.
File "tools/small_gt_generator.py", line 47, in main
frame = file_data['frame_idx']
TypeError: 'TFRecordDatasetV2' object is not subscriptable
The parameters I set are as follows:
tfrecords_load_dir = '/home/data/waymo/validation'
val_list_load_pathname = '/home/data/waymo/val.txt'
gt_load_pathname = 'gt.bin' # which is produced by gt_generator.py
gt_small_save_pathname = 'small_gt.bin'
Am I misunderstanding something?
Hi, this repo looks really nice!
I am wondering if you have any statistics of training a 3d detector on this converted kitti. Have you got some reasonable number (like >40 for vehicle on Waymo). I have tried some other third-party converters and wrote some of my own, but in my setting, the performance is really bad (about 25 on Waymo...) while I can get (>50 map on nuScenes) using the same code. So I am wondering if this repo + kitti format can come to the rescue...
Best
Hi,
After I installed the tensorflow=2.1.0 and waymo-open-dataset-tf-2-1-0==1.2.0, I still get the attribute error:
print(obj.num_lidar_points_in_box)
AttributeError: num_lidar_points_in_box
and I haven't build the "Metrics Computation" and "Python Utilities" successfully, will it affect?
Hi,
Using the converted results of part of waymo, I train and test the PointRCNN project
1.training set: training_0000 and training_0001
2.testing set: training_0002
The training loss is normal, which is about 1.0~1.5, but the mAP is weird, which is < 1.0.
Considering the format of the waymo_kitti_results is the same as that of kitti, It's strange to get such a poor performance.
Hi! Using the converted camera intrinsic and 3D object detection, I calculated the projected 2D bboxes. However, it seems they are often larger than the real object, and it seems there's a distortion problem while visualization. Do you have any suggestions to handle these problems?
Hi,
I followed all your instructions, but no matter if I use
pip3 install tensorflow==2.1.0
and pip3 install waymo-open-dataset-tf-2-1-0==1.2.0 --user
or
pip3 install tensorflow==2.0.0
and pip3 install waymo-open-dataset-tf-2-0-0==1.2.0 --user
I get the error listed below.
$ python converter.py
Traceback (most recent call last):
File "converter.py", line 12, in <module>
from waymo_open_dataset.utils.frame_utils import parse_range_image_and_camera_projection
ModuleNotFoundError: No module named 'waymo_open_dataset.utils.frame_utils'
$ pip list
Package Version
--------------------------- -------------------
absl-py 0.10.0
aiohttp 3.6.2
argon2-cffi 20.1.0
astor 0.8.1
astunparse 1.6.3
async-generator 1.10
async-timeout 3.0.1
attrs 20.2.0
backcall 0.2.0
bleach 3.2.1
cachetools 4.1.1
certifi 2020.6.20
cffi 1.14.3
chardet 3.0.4
cycler 0.10.0
decorator 4.4.2
defusedxml 0.6.0
entrypoints 0.3
gast 0.2.2
google-auth 1.22.0
google-auth-oauthlib 0.4.1
google-pasta 0.2.0
grpcio 1.32.0
h5py 2.10.0
idna 2.10
importlib-metadata 2.0.0
ipykernel 5.3.4
ipython 7.18.1
ipython-genutils 0.2.0
ipywidgets 7.5.1
jedi 0.17.2
Jinja2 2.11.2
jsonschema 3.2.0
jupyter-client 6.1.7
jupyter-core 4.6.3
jupyterlab-pygments 0.1.2
Keras-Applications 1.0.8
Keras-Preprocessing 1.1.2
kiwisolver 1.2.0
Markdown 3.2.2
MarkupSafe 1.1.1
matplotlib 3.3.2
mistune 0.8.4
mpmath 1.1.0
multidict 4.7.6
nbclient 0.5.0
nbconvert 6.0.6
nbformat 5.0.7
nest-asyncio 1.4.1
nose 1.3.7
notebook 6.1.4
numpy 1.18.5
oauthlib 3.1.0
open3d 0.10.0.0
opencv-python 4.4.0.44
opt-einsum 3.3.0
packaging 20.4
pandocfilters 1.4.2
parso 0.7.1
pexpect 4.8.0
pickleshare 0.7.5
Pillow 7.2.0
pip 20.2.3
prometheus-client 0.8.0
prompt-toolkit 3.0.7
protobuf 3.13.0
ptyprocess 0.6.0
pyasn1 0.4.8
pyasn1-modules 0.2.8
pycparser 2.20
Pygments 2.7.1
pyparsing 2.4.7
pyrsistent 0.17.3
python-dateutil 2.8.1
pyzmq 19.0.2
requests 2.24.0
requests-oauthlib 1.3.0
rsa 4.6
scipy 1.4.1
Send2Trash 1.5.0
setuptools 49.6.0.post20200925
six 1.15.0
sympy 1.4
tensorboard 2.0.2
tensorboard-plugin-wit 1.7.0
tensorflow 2.0.0
tensorflow-estimator 2.0.1
tensorflow-gpu 2.0.0
termcolor 1.1.0
terminado 0.9.1
testpath 0.4.4
tornado 6.0.4
tqdm 4.50.0
traitlets 5.0.4
typing-extensions 3.7.4.3
urllib3 1.25.10
waymo-open-dataset-tf-2-0-0 1.2.0
wcwidth 0.2.5
webencodings 0.5.1
Werkzeug 1.0.1
wheel 0.35.1
widgetsnbextension 3.5.1
wrapt 1.12.1
yarl 1.6.0
zipp 3.2.0
Any suggestions on how to resolve this?
Hey all and @caizhongang
I checked the conversion data and worked on it for nearly one month. I found that data conversion is not so good for object detection. Here you can see error in the data conversion this
Before using this conversion tool, one has to remove the frames where ground truths are there even objects are invisible.
Since waymo dataset was designed for lots of purpose. we need to be careful while using waymo dataset
Hi, zhongang! Thank you for this conversion tool.
I saw that this repo can only get the calib files of the front camera, can I get the calib files of the other four cameras?
Hi, I want to reference the number in a paper. Is there a rough estimate about the inference time (with the corresponding hardware)? Also, from the paper, single model PartA^2 gets 62.36 vehicle maph on the validation set?
Hi,
there is a line in conver.py: if filter_empty_3dbox and obj.num_lidar_points_in_box < 1
And the error is as follows:
File "converter.py", line 360, in save_label
if filter_empty_3dbox and obj.num_lidar_points_in_box < 1:
AttributeError: num_lidar_points_in_box
The attribute "num_lidar_points_in_box" is not given in https://github.com/waymo-research/waymo-open-dataset/blob/master/waymo_open_dataset/label.proto
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.