render later after fonts are downloaded

active trashcan icon for labels is brighter than images

small new bug, use arrow down/up when labelling first object

when opening up the labeller modal, there is a scrollbar

the label trashcan has a stronger tint than the image

zoom fix - check how it looks after making the original image full height

add gifs

remove unused asset files

social media preview image for https://myvision.ai

check whether the hyperlink to index.html in the readme is enough

when user holds ctrl and scrolls, should zoom-in on the image (be ware of use cases where the user may be drawing or editing shapes)

the remove images modal should be smaller

potentially emit the use of the publicDev folder

potentially decrease the height of the upload datasets modal

upon viewing an image with ml shapes on it and then moving to another one, the shape's fill should disappear

think about caching some of the 'do not show again' modal data, settings

default canvas does appear to have highlight (multiple shape) capabilities, investigate

every 10 mins after the last save, should get a brief notification to remember to save the work

can't use enter in machine learning modal

the labeller overflow still does happen, check if some properties have changed

---------------------------------------------

Things to check on chrome browsers on other pcs:

weird auto horizontal rotate for some images on firefox

potential fix:
img {
    image-orientation: from-image;
}

shape labeller modal  horizontal and vertical overflow fake borders
label list dropdown horizontal and vertical overflow fake borders
machine learning generated labels horizontal and vertical overflow fake borders
make sure the dim effect is not choppy for slower pcs

-------------------------------------------------

UX:

concider the fact that after editing a label in continuous drawing mode, we should maybe go back to the drawing mode

check whether it is more approriate to add bounding boxes by holding key or clicking it multiple times

highlight shapes to indicate that the user needs to click on them to start adding/removing points - check if really needed

Check to see if the upload images and remove image button are not too high up, causing the user to click remove image when
their actual intention is to remove a shape. To fix this, can potentially move the buttons lower by increasing their padding top
or highlight the image to be removed (in red) on remove image button hover (border or potentially overlay)
Check to see if the new border functionality is not causing users to think that on hovering the delete shape button, the shape would
be highlighted

check if export datasets is enough of an indicator that it is a download functionality
(reason not using download is because the word is long for the popup)

Or maybe having the 'Download Datasets' popover and Export button is clear enough

Change the HTML naming conventions with the final decisions!!!!

--------------------------------------------------

the current tfjs library version displays a lot of warnings in the console, look out for a new tfjs version to remove them (which supports source maps)

the word "current" is used heavily within this project and should be replaced with "active"

ideally should write new code in a separate module to enable code splitting into separate chunks (check is sharing is required)

the dimensions used for recording examples were 1920x822

The export datasets button is not disabled on load time to enable the user explore the formats available

encode all non-alphanumeric symbols in index into entity code

Decided not to have a continue to edit polygon button on the popup as it made it look vulgar and does is not a functionality that users
would really want from the get-go as the rest of the polygon edit tools would enable them to ammend their polygons post creation

Decided not to use image segmentation because it works on pixel bases, where each and every boundary is drawn by an array of pixels
However, can be achieved by generating an algorithm that will draw polygons based on the boundaries found on the pixels variable's value
on line 318, file called as segment-annotator.js.

https://github.com/kyamagu/js-segment-annotator

The following repo does use fabricjs, but to only draw a line;
https://github.com/AKSHAYUBHAT/ImageSegmentation

The nice to haves:

when a user tries to drop an image on top of the app screen, it should automatically trigger uplaod (or in the case of uploading datasets - add to the mdodal)
https://webpack.js.org/guides/caching/#extracting-boilerplate

split up the result bundle into chunks to increase download

Use the following to identify the browser:
https://github.com/WhichBrowser/Parser-JavaScript?fbclid=IwAR0pxGzSJEdjUKMkWRryUI8N5c81h5YyiK1fVXCK_IamftA_vCpD21DygyM

select multiple images or shapes to delete (via ctrl)

can update the image list thumbnails by taking screenshot from farbicjs canvas when something changes

the label should not track off the screen

polygon size reduction when zoomed in is a little bit off

I believe the bounding box drawing bottom edge is not working in horizontally narrow images going by the trend

control the darkness of the background

control the thickness of shape corner sizes

control the size of labels

change label colors

rebind keys

export custom settings

animations for image upload, new label create

should alert the user when unsaved changes before leaving (potentially too intrusive for new people)


# For more information see: https://help.github.com/actions/language-and-framework-guides/publishing-nodejs-packages
