Coder Social home page Coder Social logo

changsin / adart Goto Github PK

View Code? Open in Web Editor NEW
0.0 0.0 2.0 152.02 MB

Adart is a data reviewing tool written in Streamlit and ReactJS.

Python 79.10% HTML 5.26% CSS 0.16% TypeScript 10.20% Jupyter Notebook 4.88% Mako 0.08% Shell 0.15% Dockerfile 0.16%
ai data streamlit

adart's People

Contributors

changsin avatar ysjadhav avatar

Watchers

 avatar  avatar

adart's Issues

Spaces in Project Name

If there is a space in the name of the Project (while creating the project), the images and image data doesn't get loaded.
The label data is loaded as normal

Spaces in the Task Name don't matter

Create task subfolders under each project folder

For each task, create a subfolder.
Right now, all data go into a common folder called "data"
This creates a problem when there are duplicate filenames.
To prevent such problems, we should create a subfolder for each task and then create the data subfolder.

A consequence that we need to deal with is how to glob all data files when doing reporting. This should not be a problem since you can add all data from all tasks by globbing each sub-task folder.

Refactor file converter code in tasks.py

Here is the code that does the file format conversions:

adart/src/pages/tasks.py

Lines 206 to 212 in 5f9d60e

for idx, anno_filename in enumerate(saved_anno_filenames):
if labels_format_type == CVAT_XML:
converted_filename = _convert_anno_files(labels_format_type,
save_folder,
saved_data_filenames,
[anno_filename])
converted_filenames.append(converted_filename)

We would like to achieve two things:

  1. Inside _convert_anno_files(), label files are converted into Adart data_labels format in two steps: first into Adq format and then into Adart data_labels format. We want to do it in one direct step.
  2. Refact _convert_anno_files() to have a streamlined workflow so that later we can have something like

reader = BaseReader()
// instantiate reader with the correct reader type

data_labels = reader.parse(...)
data_labels.save(converted_filename)

Viewer: Invalid value 1794 on selecting a boundary

To repro:

  1. Load up the sample StVision file
  2. Click on the border right beside the VP (vanishing point) in the image.
Screenshot 2023-05-14 at 8 06 32 AM

Expect:
The attributes of the object should appear below.

Actual:
The thumbnail view appears but an error occurs:

File "/Users/changsin/workspace/adart/src/viewer/app.py", line 224, in main
_display_type_attributes(selected_shape, key=key)
File "/Users/changsin/workspace/adart/src/viewer/app.py", line 67, in _display_type_attributes
index=Type3PositionE.get_index(type3value) if type3value else 0)
File "/Users/changsin/workspace/adart/src/common/constants.py", line 164, in get_index
for i, member in enumerate(Type3PositionE):
ValueError: Invalid value 1794

It seems that 1794 is a value that needs to be added to Type3PositionE Enum

Image/Annotation/Error logging error + LAG

ISSUE 1:
There is a new issue with the way the errors are getting logged.
(Could not repro when a new folder was uploaded)

In the BBox project, the image annotation file for the 0002.jpg image was re-written with the annotation file information for the 0001.jpg image.
During task review, the first image and the annotations are loaded in normally.
However, the first image and it's annotations are now being shown in place of the second image and the corresponding annotations as well.
Images ( and labels) third image onwards are normal.

Somehow the file itself is being overwritten.

ISSUE 2:
In addition to this, there is a lag in the loading of the images.
When you open up a folder to review in the Review Task option, the image flashes on the screen and then disappears. You have to go to the next image and then come back to see the image and interact with it.

Similarly there is a slight lag in the loading of the image (and the attributes) when you use the arrow buttons or even through the drop down menu and want to open another image to review.

image
image

Delete button to take effect immediately

Repro steps:

  1. Create a project and a task with some images
  2. Add a label by clicking 'Untagged' button in the viewer
  3. Press "Delete" button

Expect:
The label is to be deleted immediately

Result:
It is not removed. But if you move Next or Previous and come back to the original image, it is gone.

Reviewer category: Review task bug

When you are in the reviewer mode and go to select a task to review the task, you can select the project, but the dropdown menu is blank for the task selection and you can't actually review any tasks.

Could be related to the permissions that a reviewer has?

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.