bernwang / latte Goto Github PK

Hi,

I'm trying to annotate my own dataset. I'v already had my pointclouds, no_ground_pointclouds and images ready, but I got the following error when I ran app.py:

/app/test_dataset/0_drive_0064_sync/oxts/32C_000000.txt not found.

What are the oxts files and how to generate it? Btw, the data are not loaded if I name the folder myself, and I can only load if I move the data into one of the existing dataset folders.

Any help would be nice,
Thanks

I installed latte with Python 3.7.15 with conda, because >=3.8 is not working with versions of matplotlib and tenserflow in requirenment file.
I do not see any errors, but buttons with Mask RCNN Labels and Object from Image labels are not showing anything.

Did somebody else had this issue? @alvinwan @bernwang

I run the code on ubuntu16.04 in a server, but I can't connect to the webpage of IP:5000, and I can't find the code relate to the port configure. Could you please tell me where to change the port?

How to use custom data png files for the annotation

I would like to get some help to enable bounding box height adjustment.
The bounding box height is not accurate. However, the bounding box height is not adjustable in current version.

Dear Bernwang,

The tool works flawless, Thank you very much for your work!
I am wondering if it is possible to add a functionality to do 3-D bounding box annotation? Specifically, the ability to specify the ground level and the height of the object?

hello,i used ubuntu16.04 system,and i can open it in firefox,but i can not draw the boxes on this browser.So i changed another browser,edge.I can draw the boxes,but i can not save it.It is so strange.I wonder if it is because my browser?

How can we convert PCD or rosbag to the format this tool need?

I would like to use your annotator. However, I am working only on point cloud. Could you advice me on how to remove the MaskRCNN and also oxts/ loading process so that the annotator still have the labeling functionality, one-click bounding box and also tracking?

As the reason you know, I can't access google, and after running the latte, and I get the bug "GET https://ajax.googleapis.com/ajax/libs/jquery/3.3.1/jquery.min.js net::ERR_CONNECTION_RESET".
and solve this problem by change this line

Hallo Sir...

Thank you for the Useful 3d Bounding Box Automatic Labelling... Now I am using Latte for Automatically Labelling 3D Bounding Box in LiDAR Data...

By the way I want to confirm there is no Z Coordinates and Heigh in Bounding Box ? Because as I know in 3 D Bounding Box there are Z Coordinates and Height for Cars and Vehicles Object Detection...

I'm currently using ubuntu18 and chrome browser. I installed all dependencies as instructed and also checked both master and evaluation branch. The mask-rcnn label and cropped image cannot be loaded.

I found that index.html in your app/templates. But when I tried to open it in my chrome and firefox, it shows below.

I am using ubuntu 16.04... Can you help me? I really want to use your annotator

Hi, I tried testing with one of the KITTI rectified-processed datasets but was stuck because of file structure mismatch. What tools did you use to extract non-ground related point cloud data?

I'm interested in the tracking feature you proposed for annotating sequential frames. However, I noticed there's a tracker.py and inside bounding_box_predictor.py. there's also a function predict_next_frame_bbox. So which one implements the tracking feature? Or what's the difference between them?

And I notice there's a paticle_filter.py but never used. Can you please help explaining it a little bit? Thanks!

1. when I run your project on ubuntu16.04 and chrome, I can't see the Mask RCNN Labels and Object from image, Do you know why?
2. what algorithms do you use to remove ground?
3. in your test_data directorys, images have only one view, but I think there is supposed to be 360 perspective, 6 cameras, 6 images every frame.
   4） when i switch to next frame, the id for one car will change, the track performance is not very good. Is there any mistakes about my operations or sth else?

Hi there can someone tell me on how do i add multiple properties in each label
For example under the car you need to choose if its occluded or not by how many percentage.
and under the car label there is also a category on what kind of car is it.

app/static/js/box.js filename differ from follow:

Hi,
Thanks for the tool.
Could you tell me how the labels to the point clouds are saved?

Based on the demo video

https://www.youtube.com/watch?v=QWjWpqvYA_c&feature=youtu.be

There is a feature to integrate segmentation masks to improve bounding box drawing, but I could not see any readme / guide to explain how to achieve this.

hello,i used Ubuntu 16.04 ,when i python app.py, i opened the http://127.0.0.1:5000/ in firefox, but only displayed some words, no picture was displayed, and then i changed the browser ,use chrome, but the same situation occurred.

The following is the run app.py displayed:

127.0.0.1 - - [15/Jul/2019 09:02:58] "GET / HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/styles/style.css HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/styles/imagePanels.css HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/js/libs/three.min.js HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/js/init.js?nocache HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/js/frame.js?nocache HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/js/libs/OrbitControls.js?nocache HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/js/libs/Detector.js HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/js/libs/stats.min.js HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/js/libs/FileSaver.min.js HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/js/libs/dat.gui.min.js HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/js/box.js?nocache HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/js/app.js?nocache HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/js/utils.js?nocache HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/js/pointcloud.js?nocache HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/js/ObjectTable.js?nocache HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/js/Controls.js?nocache HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/js/Evaluator.js?nocache HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/js/Box.js?nocache HTTP/1.1" 404 -
127.0.0.1 - - [15/Jul/2019 09:03:00] "GET /static/js/main.js?nocache HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:03:00] "GET /static/js/imagePanels.js?nocache HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:02:59] "GET /static/js/Evaluation.js?nocache HTTP/1.1" 200 -
127.0.0.1 - - [15/Jul/2019 09:03:21] "GET /static/js/Box.js?nocache HTTP/1.1" 404 -
127.0.0.1 - - [15/Jul/2019 09:03:23] "GET /favicon.ico HTTP/1.1" 404 -

While executing the command pip3 install -r requirements.txt, I'm getting following errors python setup.py egg_info did not run successfully. RuntimeError: Python version 2.7 or 3.4+ is required. I'm using Python version 3.10.6.

i have .laz and .laz format for lidar data. can i used this format? if not then is there any way to convert this format to .bin?

Hello, I need 3D label, but I don't know how to label the height of the object. Can you help me?

Hi,

1. I see bounding boxes in 2D only.
2. I cannot see the image itself in the left part of the screen (as seen in the YouTube video).
3. Save button isn't working - the file in the oxt directory isn't being changed after adding a bounding box, for example.

Thanks!