lacmus-foundation / lacmus-app Goto Github PK

We need to create control where it is possilbe to see current events -- warn messages, error, info . Just some TextBox or similar control where logs are collected for current user session. It doesn't exclude file logs which are persisted among all user sessions and probably should be more verbose -- contain debug messages.

• next photo (ctrl+pageDown)
• prev photo (ctrl+pageUp)
• show detect Box (example space)
• zoom (+ or -)
• show full image (autozoom for full work area) (example ctrl+0)
• move image (arrows)

Thumbnails of Photos are stored in the standard ListBox and are displayed all at once.
Avalonia is not able to discard more than 10,000 elements at a time. For this reason, the application freezes if the user uploads more than 10,000 photos.

It is necessary to display the photographs in a box box in portions - that is, to make the list pagination.
Here's how it goes in the WPF:

• https://github.com/Dessembrae/MVVM-ListView-Pagination
• https://github.com/SuperJMN/XAMLPagingControl
• https://www.powerobjects.com/blog/2013/02/19/wpf-paging-control-wpf-applications/
• use google: "wpf pagging control", "wpf listbox pagination", wpf list pagination" ... 😄

When uploading new photos or switching between pages or completing a prediction using a neural network, the selected photo is shown in the photo collection, but there is no full picture in the preview. You have to click on the next photo and go back.

This is because the update ui function is called and does not redraw the selected photo. It is necessary to track this moment.

Actions:

• select "show photos with object" filter
• chose "file" -> "import from XML..."
• see the bug: collection is not updated. The problem is reproduced on versions: 0.3.3.preview [0, 1, 2, 3]

Actions:
In show metadata window add a point to google and Yandex map when they open. Now on the map only the desired area is opened but a specific point is not shown. It is not comfortable.

In the Docker class, implement display of the container loading status and display it in the console at first and later in the status bar of the application.

Use sharp-in technology via ML.NET and OPEN VINO tookit to improve image quality

Add global application settings.

The application needs settings where the user can select and configure a number of functions such as:

• choose a theme
• choose the option of using a neural network (cpu / gpu / remote server)
• Ui language
• it is possible to add more functions

Note1: It is necessary to think and add here those functions that the user can configure.

Note2: We should add this option after refactoring the code.

If we stay inin the “Photos with objects” filtering mode, and then start processing the next batch of photos, then upon completion this list is not updated, you need to manually enable the display of all photos and then return to the filter.

In short, the filter is not automatically applied when updating the original image collection. In order for the filter to be updated, you need to call it explicitly.

Действия:

• запустить приложение
• нажать кнопку Open, выбрать папку с фотографиями; в левой части отображаются фото
• нажать на первую фотографию в левой части (одинарный или двойной клик)

Выбранная фотография не отображается в правой части

Примечание:

• Проблема существует, только если выбирать в самом начале первое фото
• Если после выбора первого фото выбрать друге фото, то другое фото отобразится как нужно. После этого действия выбор первого фото приводит к нормальному поведению (фото отображается)

For example DJI {~} 6.JPG omitted a significant part of the name. There is more than enough space for a full name (less than half of the available space is used).

And in general, the available space is not used optimally: images can be much larger, the active element can be highlighted with the color of the entire background, and not take away 8px border.

Move "Bug report" to "image" menue section

Avalonia ui developers recently released new themes. It all looks pretty cool. You can add a few to the project after refactoring the code.

The source link:
https://m.habr.com/ru/post/487000/

Thanks @worldbeater for awsone skines

cann not open url at windows 10

I have already tried several times to solve this problem and for some reason, somewhere, opening links still does not work. Personally, on my test machine with Windows everything works fine. It is obvious that it is tedious to carry out additional testing of this function.

[02:28:47 ERR] Unable to ope url https://yandex.ru/maps/?ll=37.788036%2C55.91099&z=15&mode=whatshere&whatshere%5Bpoint%5D=37.788036%2C55.91099&whatshere%5Bzoom%5D=15.
System.ComponentModel.Win32Exception (2): Не удается найти указанный файл.
   at System.Diagnostics.Process.StartWithCreateProcess(ProcessStartInfo startInfo)
   at System.Diagnostics.Process.Start()
   at System.Diagnostics.Process.Start(ProcessStartInfo startInfo)
   at System.Diagnostics.Process.Start(String fileName)
   at LacmusApp.ViewModels.MetadataViewModel.OpenUrl(String url) in /home/gosha20777/files/projects/lacmus-app/src/ViewModels/MetadataViewModel.cs:line 119

Functionality for canceling long-running operations should be introduced. For example, loading 200 img files(gpa_01) took more than 1 minute. Long-running operations should be cancelable -- predict all, open and so on...

Unable to create docker container: invalid id Error response from daemon: Unknown runtime specified nvidia
GPU ml model dose not worked.
Actions:

• use mlConfig-gpu.json
• run the application and download the ml model
• click Predict All

Get an error:

[13:56:21 ERR] Unable to get prediction.
System.Exception: Unable to initialize ml model
---> System.Exception: Unable to create docker container: invalid id Error response from daemon: Unknown runtime specified nvidia
   at RescuerLaApp.Models.Docker.Docker.CreateContainer(IDockerImage image) in /home/gosha20777/files/projects/lacmus-app/Models/Docker/Docker.cs:line 168
   at RescuerLaApp.Models.ML.MLModel.Init() in /home/gosha20777/files/projects/lacmus-app/Models/ML/MLModel.cs:line 44
   --- End of inner exception stack trace ---
   at RescuerLaApp.Models.ML.MLModel.Init() in /home/gosha20777/files/projects/lacmus-app/Models/ML/MLModel.cs:line 65
   at RescuerLaApp.ViewModels.MainWindowViewModel.PredictAll() in /home/gosha20777/files/projects/lacmus-app/ViewModels/MainWindowViewModel.cs:line 404

The following settings are required

• configuration settings of the ml model
  • ml model configuration
  • ml model manage (download/install/remove versions)
  • ml model repositories configuration
• color scheme settings
• interface language settings

On my machine, win 8.1, buttons doesn't react on click, nothing happens.

Add the logger with file and console handlers instad of Console.Writeline() method.

Currently, aboutiinformation is only displayed using the messagebox window. Lets make it more nice!

Avalonia ui added svg support. We can make the application information window more similar to what it looks like on Mac OS / ubuntu / Windows.
We can also add a donate button;-)

Update developer documentation and bring it up to date (in Russian).
The Page link
Add the following sections to the page:

• building and dependenses
• class diagram (after refactoring)
• how to contribute

Maybe something esle...

use https://github.com/wieslawsoltes/Dock to make more flexible and convenient interface

https://github.com/wieslawsoltes/Dock/wiki/Getting-Started - useful link

We need support for several languages ​​in the application interface: Russian and English.

in the program settings window, the parameters are always returned to their default state

If you click "Open" and select "cancel" in the folder selection dialog box, the current list of photos will be cleared.

In other words, if the dialog box returns an empty list, then the current collection is overwritten.

Hello!

We can upgrade TargetFrameworks to netcoreapp3.1?

In addition, it would be better if everything except a few files (.gitignore, CODE_OF_CONDUCT.md, CONTRIBUTING.md, LICENSE, README.md, RescuerLaApp.sln, mk-release.sh, nuget.config) was in the src / source folder for clarity :)

Show 0 of 2 and 1 of 2. Require increment number

When the program is running, the gru or cpu are utilized jerkily - (0 - 100%).
This is due to the fact that during the loading of data and the formation of the request, the resources brighten.

It is necessary to analyze 2 facts

1. How forte byte array as base64 strings affects performance.
2. It is likely to load the photo into memory asynchronously until it is sent to the neural network for processing.

Open source should be free! To hell with passwords and restrictions.

P.s. Authorization was introduced first at the request of Lisa Alert. However, then after much persuasion and explanation, we convinced them that there was no point in authorization.

Users need a simpler way to manage and use the application.

I propose to make a wizard for detecting objects according to what Windows OS and its method of installing programs use for us.

The master plans 3 steps.

1. Set input data.
2. Set the place where to save the processed data.
3. Configure the neural network model.

After that a status bar should be displayed and some statistics of the prediction process should be written.

At the end, export the final statistics and offer to close the window or repeat the operation.

This window should work on top of the main window and duplicate its functionality. The user is free at any time to close the wizard window and continue working in the main window.

Below is a diagram of how this might look.

No allowable limits are specified when moving / scaling an image. For this reason, photography can run away. This problem is especially acute when using the laptop touchpad.

You must set limits to move or scale the image.

To display a picture on the screen, we use a pan and zoom controller by @wieslawsoltes

Currently, metadata is only displayed using the messagebox window. However, it is necessary:

• make a friendlier interface
• put this code in a separate vm
• add the function "copy coordinates to the clipboard" so that later they can be pasted into Google maps for example
• add a tabular view for this data

It is necessary to automate the collection of false positives and people not found in the lacmus photo storage.

It will be a plus to give the opportunity to mark up directly in the application.

Нужно автоматизировать сбор ложных срабатываний и не найденных людей в хранилище лакмуса.

Будет плюсом дать возможность размечать прямо в приложении.

• run the application
• upload a photo
• select photo
• select menu Image -> Predict This
  Search for an object in the photo does not occur. In the log - no mention of the beginning of the prediction.

Add user documentation to the wiki of the Lacmus project in Russian.

Добавить документацию следующей структуры

• заглавная страница - краткое описание приложения и его компонент. Снимки экрана ссылки на подразделы ссылки на донат нам на хлебушек и пивас.
  • системные требования (готово)
  • процедура установки - пошаговая инструкция установки приложения на поддерживаемые платформы: Linux (debian based), OS X, `Windows (отдельно для Windows 10 pro, (с использованием docker desktop) и для Windows 10/8/7 с использованием docker tools)
  • обзор интерфейса - документация по использованию
  • информация о том как обновить ПО
  • информация о том как помочь проекту
    • как отправить bug report
    • как отправить нам данные
    • информация как задонатить в проект

Примечание: будет здорово везти документацию в стиле вопрос - ответ подобно тому как это сделано во freeBSD

It doesn’t hurt to add parameters of format-modifying output files:

• Generate XML.
• Overlay frames indicating areas of detection on photographs.
• Cut detection zones as small images (for example, 500x500 pixels).

Действия:

• Запустить Wizard, пройти все шаги. На одном из шагов выбрать место для сохранения фото
• Закрыть приложение и запустить заново
• Выбрать меню File - Import all from XML..., выбрать папку, в которых ранее сохранились фото

Фото загрузились, но отметки о найденных людях на фото не отображаются.
Включение/выключение кнопки Border ничего не изменяет

You must add the ability to filter the list of photos

• show only photos with objects marked with a neural network
• show photos marked as favorites.
• show all photos.

To do this, use reactiveui dynamic data package

• https://habr.com/ru/post/445098/
• https://github.com/reactiveui/DynamicData/wiki/Introduction-for-ReactiveUI-users
• https://github.com/RolandPheasant/DynamicData.Snippets - code examples

Unit Tests? No, not heard 😃

It is necessary to separate the logic of some parts of the application and refactor the application code. Here are the key issues in the code:

• The MainViewModel.cs class is overloaded with logic. It is necessary to divide this class into several classes and transfer the logic to other classes.
• it is necessary to remove the logic of loading the neural network model from the Docker class and share its responsibilities
• It is necessary to process the class of the neural model and implement its configuration using the configuration file (for example, which version to use, cpu, gpu how to call the neural network model using docker locally or using a remote server)
• Rewrite code as recommended by ReactiveUI
• Identify other defects and fix them.

The main goal of refactoring is to provide a higher level of modularity of the application and allow you to quickly implement new functions and look for bugs.

Note: for a number of iconic points in the code refactoring, individual issues will be entered.

Add the following blocks to the readme file:

• brief description of the application
• screenshot
• links to user documentation
• links to development documentation
• information on how to compile the application from source
• release notes
• license information

Key page Down (page Up) moved to next (prev) bbox in image (or next (prev) image) from zoom 1:1

Цель этого и #115, #114 -- упростить перегруженное окно менеджера моделей.