Angela  Dickens

Angela Dickens

1597251300

Deep Learning Image Classification with Fastai

TL;DR

If you are beginning to feel burnt out on learning a subject it is often beneficial to take a step out of the weeds. I like to build something fun and easy to regain positive momentum on my learning journey. This is one project that helped get my creative juices flowing again. Load my notebook from my GitHub repository into Google Colab and upload the Kaggle data set to learn how to build an image classifier using the fastai software! Don’t forget to set the hardware accelerator to GPU!

P.S. Directly upload notebook to Colab by going to File →Upload Notebook → GitHub Tab→BSamaha/Chest-Xray-FastAI

Image for post

Image for post

How to easily upload a GitHub Notebook to Google Colab


Intro

The Dunning-Kruger Effect

You have set a goal such as wanting to be a data scientist or data engineer. Your passion and energy are through the roof as you rifle through all the material on the subject you can find. Perhaps you even started a coding boot camp or a Coursera class to be your guide on your path to data mastery. Your brain becomes saturated with dopamine as you quickly rack up quick wins. First, it was basic python programming. Then you made your first linear regression with sci-kit learn — machine learning isn’t so bad after all! You quickly absorbed all information thrown at you. Nothing can stop you.

I’ve been there, and I still go there. I am human after all. I’m referring to the peak of “Mt. Stupid” in The Dunning-Kruger Effect. Linear regression was easy with a perfectly curated data set, but that’s not reality. Finally, after about a month or two of flying through the material with ease, you hit a wall named statistics. You grind through it and barely make it out — except you feel like you haven’t retained anything. Do people remember all this distribution stuff? Spoiler Alert: yes, and you will be questioned about them. As you pursue on you realize your python skills are nowhere near where they should be. You decide to look at a couple of job postings and then it hits you. You don’t have an advanced graduate degree in computer science, and you don’t know how to pronounce some requirements of the job much less know what they are — Hadoop? Welcome to the Valley of Despair. You want to quit and move on with your life.

#deep-learning #machine-learning #deep learning

What is GEEK

Buddha Community

Deep Learning Image Classification with Fastai
Queenie  Davis

Queenie Davis

1653123600

EasyMDE: Simple, Beautiful and Embeddable JavaScript Markdown Editor

EasyMDE - Markdown Editor 

This repository is a fork of SimpleMDE, made by Sparksuite. Go to the dedicated section for more information.

A drop-in JavaScript text area replacement for writing beautiful and understandable Markdown. EasyMDE allows users who may be less experienced with Markdown to use familiar toolbar buttons and shortcuts.

In addition, the syntax is rendered while editing to clearly show the expected result. Headings are larger, emphasized words are italicized, links are underlined, etc.

EasyMDE also features both built-in auto saving and spell checking. The editor is entirely customizable, from theming to toolbar buttons and javascript hooks.

Try the demo

Preview

Quick access

Install EasyMDE

Via npm:

npm install easymde

Via the UNPKG CDN:

<link rel="stylesheet" href="https://unpkg.com/easymde/dist/easymde.min.css">
<script src="https://unpkg.com/easymde/dist/easymde.min.js"></script>

Or jsDelivr:

<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/easymde/dist/easymde.min.css">
<script src="https://cdn.jsdelivr.net/npm/easymde/dist/easymde.min.js"></script>

How to use

Loading the editor

After installing and/or importing the module, you can load EasyMDE onto the first textarea element on the web page:

<textarea></textarea>
<script>
const easyMDE = new EasyMDE();
</script>

Alternatively you can select a specific textarea, via JavaScript:

<textarea id="my-text-area"></textarea>
<script>
const easyMDE = new EasyMDE({element: document.getElementById('my-text-area')});
</script>

Editor functions

Use easyMDE.value() to get the content of the editor:

<script>
easyMDE.value();
</script>

Use easyMDE.value(val) to set the content of the editor:

<script>
easyMDE.value('New input for **EasyMDE**');
</script>

Configuration

Options list

  • autoDownloadFontAwesome: If set to true, force downloads Font Awesome (used for icons). If set to false, prevents downloading. Defaults to undefined, which will intelligently check whether Font Awesome has already been included, then download accordingly.
  • autofocus: If set to true, focuses the editor automatically. Defaults to false.
  • autosave: Saves the text that's being written and will load it back in the future. It will forget the text when the form it's contained in is submitted.
    • enabled: If set to true, saves the text automatically. Defaults to false.
    • delay: Delay between saves, in milliseconds. Defaults to 10000 (10 seconds).
    • submit_delay: Delay before assuming that submit of the form failed and saving the text, in milliseconds. Defaults to autosave.delay or 10000 (10 seconds).
    • uniqueId: You must set a unique string identifier so that EasyMDE can autosave. Something that separates this from other instances of EasyMDE elsewhere on your website.
    • timeFormat: Set DateTimeFormat. More information see DateTimeFormat instances. Default locale: en-US, format: hour:minute.
    • text: Set text for autosave.
  • autoRefresh: Useful, when initializing the editor in a hidden DOM node. If set to { delay: 300 }, it will check every 300 ms if the editor is visible and if positive, call CodeMirror's refresh().
  • blockStyles: Customize how certain buttons that style blocks of text behave.
    • bold: Can be set to ** or __. Defaults to **.
    • code: Can be set to ``` or ~~~. Defaults to ```.
    • italic: Can be set to * or _. Defaults to *.
  • unorderedListStyle: can be *, - or +. Defaults to *.
  • scrollbarStyle: Chooses a scrollbar implementation. The default is "native", showing native scrollbars. The core library also provides the "null" style, which completely hides the scrollbars. Addons can implement additional scrollbar models.
  • element: The DOM element for the textarea element to use. Defaults to the first textarea element on the page.
  • forceSync: If set to true, force text changes made in EasyMDE to be immediately stored in original text area. Defaults to false.
  • hideIcons: An array of icon names to hide. Can be used to hide specific icons shown by default without completely customizing the toolbar.
  • indentWithTabs: If set to false, indent using spaces instead of tabs. Defaults to true.
  • initialValue: If set, will customize the initial value of the editor.
  • previewImagesInEditor: - EasyMDE will show preview of images, false by default, preview for images will appear only for images on separate lines.
  • imagesPreviewHandler: - A custom function for handling the preview of images. Takes the parsed string between the parantheses of the image markdown ![]( ) as argument and returns a string that serves as the src attribute of the <img> tag in the preview. Enables dynamic previewing of images in the frontend without having to upload them to a server, allows copy-pasting of images to the editor with preview.
  • insertTexts: Customize how certain buttons that insert text behave. Takes an array with two elements. The first element will be the text inserted before the cursor or highlight, and the second element will be inserted after. For example, this is the default link value: ["[", "](http://)"].
    • horizontalRule
    • image
    • link
    • table
  • lineNumbers: If set to true, enables line numbers in the editor.
  • lineWrapping: If set to false, disable line wrapping. Defaults to true.
  • minHeight: Sets the minimum height for the composition area, before it starts auto-growing. Should be a string containing a valid CSS value like "500px". Defaults to "300px".
  • maxHeight: Sets fixed height for the composition area. minHeight option will be ignored. Should be a string containing a valid CSS value like "500px". Defaults to undefined.
  • onToggleFullScreen: A function that gets called when the editor's full screen mode is toggled. The function will be passed a boolean as parameter, true when the editor is currently going into full screen mode, or false.
  • parsingConfig: Adjust settings for parsing the Markdown during editing (not previewing).
    • allowAtxHeaderWithoutSpace: If set to true, will render headers without a space after the #. Defaults to false.
    • strikethrough: If set to false, will not process GFM strikethrough syntax. Defaults to true.
    • underscoresBreakWords: If set to true, let underscores be a delimiter for separating words. Defaults to false.
  • overlayMode: Pass a custom codemirror overlay mode to parse and style the Markdown during editing.
    • mode: A codemirror mode object.
    • combine: If set to false, will replace CSS classes returned by the default Markdown mode. Otherwise the classes returned by the custom mode will be combined with the classes returned by the default mode. Defaults to true.
  • placeholder: If set, displays a custom placeholder message.
  • previewClass: A string or array of strings that will be applied to the preview screen when activated. Defaults to "editor-preview".
  • previewRender: Custom function for parsing the plaintext Markdown and returning HTML. Used when user previews.
  • promptURLs: If set to true, a JS alert window appears asking for the link or image URL. Defaults to false.
  • promptTexts: Customize the text used to prompt for URLs.
    • image: The text to use when prompting for an image's URL. Defaults to URL of the image:.
    • link: The text to use when prompting for a link's URL. Defaults to URL for the link:.
  • uploadImage: If set to true, enables the image upload functionality, which can be triggered by drag and drop, copy-paste and through the browse-file window (opened when the user click on the upload-image icon). Defaults to false.
  • imageMaxSize: Maximum image size in bytes, checked before upload (note: never trust client, always check the image size at server-side). Defaults to 1024 * 1024 * 2 (2 MB).
  • imageAccept: A comma-separated list of mime-types used to check image type before upload (note: never trust client, always check file types at server-side). Defaults to image/png, image/jpeg.
  • imageUploadFunction: A custom function for handling the image upload. Using this function will render the options imageMaxSize, imageAccept, imageUploadEndpoint and imageCSRFToken ineffective.
    • The function gets a file and onSuccess and onError callback functions as parameters. onSuccess(imageUrl: string) and onError(errorMessage: string)
  • imageUploadEndpoint: The endpoint where the images data will be sent, via an asynchronous POST request. The server is supposed to save this image, and return a JSON response.
    • if the request was successfully processed (HTTP 200 OK): {"data": {"filePath": "<filePath>"}} where filePath is the path of the image (absolute if imagePathAbsolute is set to true, relative if otherwise);
    • otherwise: {"error": "<errorCode>"}, where errorCode can be noFileGiven (HTTP 400 Bad Request), typeNotAllowed (HTTP 415 Unsupported Media Type), fileTooLarge (HTTP 413 Payload Too Large) or importError (see errorMessages below). If errorCode is not one of the errorMessages, it is alerted unchanged to the user. This allows for server-side error messages. No default value.
  • imagePathAbsolute: If set to true, will treat imageUrl from imageUploadFunction and filePath returned from imageUploadEndpoint as an absolute rather than relative path, i.e. not prepend window.location.origin to it.
  • imageCSRFToken: CSRF token to include with AJAX call to upload image. For various instances like Django, Spring and Laravel.
  • imageCSRFName: CSRF token filed name to include with AJAX call to upload image, applied when imageCSRFToken has value, defaults to csrfmiddlewaretoken.
  • imageCSRFHeader: If set to true, passing CSRF token via header. Defaults to false, which pass CSRF through request body.
  • imageTexts: Texts displayed to the user (mainly on the status bar) for the import image feature, where #image_name#, #image_size# and #image_max_size# will replaced by their respective values, that can be used for customization or internationalization:
    • sbInit: Status message displayed initially if uploadImage is set to true. Defaults to Attach files by drag and dropping or pasting from clipboard..
    • sbOnDragEnter: Status message displayed when the user drags a file to the text area. Defaults to Drop image to upload it..
    • sbOnDrop: Status message displayed when the user drops a file in the text area. Defaults to Uploading images #images_names#.
    • sbProgress: Status message displayed to show uploading progress. Defaults to Uploading #file_name#: #progress#%.
    • sbOnUploaded: Status message displayed when the image has been uploaded. Defaults to Uploaded #image_name#.
    • sizeUnits: A comma-separated list of units used to display messages with human-readable file sizes. Defaults to B, KB, MB (example: 218 KB). You can use B,KB,MB instead if you prefer without whitespaces (218KB).
  • errorMessages: Errors displayed to the user, using the errorCallback option, where #image_name#, #image_size# and #image_max_size# will replaced by their respective values, that can be used for customization or internationalization:
    • noFileGiven: The server did not receive any file from the user. Defaults to You must select a file..
    • typeNotAllowed: The user send a file type which doesn't match the imageAccept list, or the server returned this error code. Defaults to This image type is not allowed..
    • fileTooLarge: The size of the image being imported is bigger than the imageMaxSize, or if the server returned this error code. Defaults to Image #image_name# is too big (#image_size#).\nMaximum file size is #image_max_size#..
    • importError: An unexpected error occurred when uploading the image. Defaults to Something went wrong when uploading the image #image_name#..
  • errorCallback: A callback function used to define how to display an error message. Defaults to (errorMessage) => alert(errorMessage).
  • renderingConfig: Adjust settings for parsing the Markdown during previewing (not editing).
    • codeSyntaxHighlighting: If set to true, will highlight using highlight.js. Defaults to false. To use this feature you must include highlight.js on your page or pass in using the hljs option. For example, include the script and the CSS files like:
      <script src="https://cdn.jsdelivr.net/highlight.js/latest/highlight.min.js"></script>
      <link rel="stylesheet" href="https://cdn.jsdelivr.net/highlight.js/latest/styles/github.min.css">
    • hljs: An injectible instance of highlight.js. If you don't want to rely on the global namespace (window.hljs), you can provide an instance here. Defaults to undefined.
    • markedOptions: Set the internal Markdown renderer's options. Other renderingConfig options will take precedence.
    • singleLineBreaks: If set to false, disable parsing GitHub Flavored Markdown (GFM) single line breaks. Defaults to true.
    • sanitizerFunction: Custom function for sanitizing the HTML output of Markdown renderer.
  • shortcuts: Keyboard shortcuts associated with this instance. Defaults to the array of shortcuts.
  • showIcons: An array of icon names to show. Can be used to show specific icons hidden by default without completely customizing the toolbar.
  • spellChecker: If set to false, disable the spell checker. Defaults to true. Optionally pass a CodeMirrorSpellChecker-compliant function.
  • inputStyle: textarea or contenteditable. Defaults to textarea for desktop and contenteditable for mobile. contenteditable option is necessary to enable nativeSpellcheck.
  • nativeSpellcheck: If set to false, disable native spell checker. Defaults to true.
  • sideBySideFullscreen: If set to false, allows side-by-side editing without going into fullscreen. Defaults to true.
  • status: If set to false, hide the status bar. Defaults to the array of built-in status bar items.
    • Optionally, you can set an array of status bar items to include, and in what order. You can even define your own custom status bar items.
  • styleSelectedText: If set to false, remove the CodeMirror-selectedtext class from selected lines. Defaults to true.
  • syncSideBySidePreviewScroll: If set to false, disable syncing scroll in side by side mode. Defaults to true.
  • tabSize: If set, customize the tab size. Defaults to 2.
  • theme: Override the theme. Defaults to easymde.
  • toolbar: If set to false, hide the toolbar. Defaults to the array of icons.
  • toolbarTips: If set to false, disable toolbar button tips. Defaults to true.
  • direction: rtl or ltr. Changes text direction to support right-to-left languages. Defaults to ltr.

Options example

Most options demonstrate the non-default behavior:

const editor = new EasyMDE({
    autofocus: true,
    autosave: {
        enabled: true,
        uniqueId: "MyUniqueID",
        delay: 1000,
        submit_delay: 5000,
        timeFormat: {
            locale: 'en-US',
            format: {
                year: 'numeric',
                month: 'long',
                day: '2-digit',
                hour: '2-digit',
                minute: '2-digit',
            },
        },
        text: "Autosaved: "
    },
    blockStyles: {
        bold: "__",
        italic: "_",
    },
    unorderedListStyle: "-",
    element: document.getElementById("MyID"),
    forceSync: true,
    hideIcons: ["guide", "heading"],
    indentWithTabs: false,
    initialValue: "Hello world!",
    insertTexts: {
        horizontalRule: ["", "\n\n-----\n\n"],
        image: ["![](http://", ")"],
        link: ["[", "](https://)"],
        table: ["", "\n\n| Column 1 | Column 2 | Column 3 |\n| -------- | -------- | -------- |\n| Text     | Text      | Text     |\n\n"],
    },
    lineWrapping: false,
    minHeight: "500px",
    parsingConfig: {
        allowAtxHeaderWithoutSpace: true,
        strikethrough: false,
        underscoresBreakWords: true,
    },
    placeholder: "Type here...",

    previewClass: "my-custom-styling",
    previewClass: ["my-custom-styling", "more-custom-styling"],

    previewRender: (plainText) => customMarkdownParser(plainText), // Returns HTML from a custom parser
    previewRender: (plainText, preview) => { // Async method
        setTimeout(() => {
            preview.innerHTML = customMarkdownParser(plainText);
        }, 250);

        return "Loading...";
    },
    promptURLs: true,
    promptTexts: {
        image: "Custom prompt for URL:",
        link: "Custom prompt for URL:",
    },
    renderingConfig: {
        singleLineBreaks: false,
        codeSyntaxHighlighting: true,
        sanitizerFunction: (renderedHTML) => {
            // Using DOMPurify and only allowing <b> tags
            return DOMPurify.sanitize(renderedHTML, {ALLOWED_TAGS: ['b']})
        },
    },
    shortcuts: {
        drawTable: "Cmd-Alt-T"
    },
    showIcons: ["code", "table"],
    spellChecker: false,
    status: false,
    status: ["autosave", "lines", "words", "cursor"], // Optional usage
    status: ["autosave", "lines", "words", "cursor", {
        className: "keystrokes",
        defaultValue: (el) => {
            el.setAttribute('data-keystrokes', 0);
        },
        onUpdate: (el) => {
            const keystrokes = Number(el.getAttribute('data-keystrokes')) + 1;
            el.innerHTML = `${keystrokes} Keystrokes`;
            el.setAttribute('data-keystrokes', keystrokes);
        },
    }], // Another optional usage, with a custom status bar item that counts keystrokes
    styleSelectedText: false,
    sideBySideFullscreen: false,
    syncSideBySidePreviewScroll: false,
    tabSize: 4,
    toolbar: false,
    toolbarTips: false,
});

Toolbar icons

Below are the built-in toolbar icons (only some of which are enabled by default), which can be reorganized however you like. "Name" is the name of the icon, referenced in the JavaScript. "Action" is either a function or a URL to open. "Class" is the class given to the icon. "Tooltip" is the small tooltip that appears via the title="" attribute. Note that shortcut hints are added automatically and reflect the specified action if it has a key bind assigned to it (i.e. with the value of action set to bold and that of tooltip set to Bold, the final text the user will see would be "Bold (Ctrl-B)").

Additionally, you can add a separator between any icons by adding "|" to the toolbar array.

NameActionTooltip
Class
boldtoggleBoldBold
fa fa-bold
italictoggleItalicItalic
fa fa-italic
strikethroughtoggleStrikethroughStrikethrough
fa fa-strikethrough
headingtoggleHeadingSmallerHeading
fa fa-header
heading-smallertoggleHeadingSmallerSmaller Heading
fa fa-header
heading-biggertoggleHeadingBiggerBigger Heading
fa fa-lg fa-header
heading-1toggleHeading1Big Heading
fa fa-header header-1
heading-2toggleHeading2Medium Heading
fa fa-header header-2
heading-3toggleHeading3Small Heading
fa fa-header header-3
codetoggleCodeBlockCode
fa fa-code
quotetoggleBlockquoteQuote
fa fa-quote-left
unordered-listtoggleUnorderedListGeneric List
fa fa-list-ul
ordered-listtoggleOrderedListNumbered List
fa fa-list-ol
clean-blockcleanBlockClean block
fa fa-eraser
linkdrawLinkCreate Link
fa fa-link
imagedrawImageInsert Image
fa fa-picture-o
tabledrawTableInsert Table
fa fa-table
horizontal-ruledrawHorizontalRuleInsert Horizontal Line
fa fa-minus
previewtogglePreviewToggle Preview
fa fa-eye no-disable
side-by-sidetoggleSideBySideToggle Side by Side
fa fa-columns no-disable no-mobile
fullscreentoggleFullScreenToggle Fullscreen
fa fa-arrows-alt no-disable no-mobile
guideThis linkMarkdown Guide
fa fa-question-circle
undoundoUndo
fa fa-undo
redoredoRedo
fa fa-redo

Toolbar customization

Customize the toolbar using the toolbar option.

Only the order of existing buttons:

const easyMDE = new EasyMDE({
    toolbar: ["bold", "italic", "heading", "|", "quote"]
});

All information and/or add your own icons

const easyMDE = new EasyMDE({
    toolbar: [
        {
            name: "bold",
            action: EasyMDE.toggleBold,
            className: "fa fa-bold",
            title: "Bold",
        },
        "italics", // shortcut to pre-made button
        {
            name: "custom",
            action: (editor) => {
                // Add your own code
            },
            className: "fa fa-star",
            title: "Custom Button",
            attributes: { // for custom attributes
                id: "custom-id",
                "data-value": "custom value" // HTML5 data-* attributes need to be enclosed in quotation marks ("") because of the dash (-) in its name.
            }
        },
        "|" // Separator
        // [, ...]
    ]
});

Put some buttons on dropdown menu

const easyMDE = new EasyMDE({
    toolbar: [{
                name: "heading",
                action: EasyMDE.toggleHeadingSmaller,
                className: "fa fa-header",
                title: "Headers",
            },
            "|",
            {
                name: "others",
                className: "fa fa-blind",
                title: "others buttons",
                children: [
                    {
                        name: "image",
                        action: EasyMDE.drawImage,
                        className: "fa fa-picture-o",
                        title: "Image",
                    },
                    {
                        name: "quote",
                        action: EasyMDE.toggleBlockquote,
                        className: "fa fa-percent",
                        title: "Quote",
                    },
                    {
                        name: "link",
                        action: EasyMDE.drawLink,
                        className: "fa fa-link",
                        title: "Link",
                    }
                ]
            },
        // [, ...]
    ]
});

Keyboard shortcuts

EasyMDE comes with an array of predefined keyboard shortcuts, but they can be altered with a configuration option. The list of default ones is as follows:

Shortcut (Windows / Linux)Shortcut (macOS)Action
Ctrl-'Cmd-'"toggleBlockquote"
Ctrl-BCmd-B"toggleBold"
Ctrl-ECmd-E"cleanBlock"
Ctrl-HCmd-H"toggleHeadingSmaller"
Ctrl-ICmd-I"toggleItalic"
Ctrl-KCmd-K"drawLink"
Ctrl-LCmd-L"toggleUnorderedList"
Ctrl-PCmd-P"togglePreview"
Ctrl-Alt-CCmd-Alt-C"toggleCodeBlock"
Ctrl-Alt-ICmd-Alt-I"drawImage"
Ctrl-Alt-LCmd-Alt-L"toggleOrderedList"
Shift-Ctrl-HShift-Cmd-H"toggleHeadingBigger"
F9F9"toggleSideBySide"
F11F11"toggleFullScreen"

Here is how you can change a few, while leaving others untouched:

const editor = new EasyMDE({
    shortcuts: {
        "toggleOrderedList": "Ctrl-Alt-K", // alter the shortcut for toggleOrderedList
        "toggleCodeBlock": null, // unbind Ctrl-Alt-C
        "drawTable": "Cmd-Alt-T", // bind Cmd-Alt-T to drawTable action, which doesn't come with a default shortcut
    }
});

Shortcuts are automatically converted between platforms. If you define a shortcut as "Cmd-B", on PC that shortcut will be changed to "Ctrl-B". Conversely, a shortcut defined as "Ctrl-B" will become "Cmd-B" for Mac users.

The list of actions that can be bound is the same as the list of built-in actions available for toolbar buttons.

Advanced use

Event handling

You can catch the following list of events: https://codemirror.net/doc/manual.html#events

const easyMDE = new EasyMDE();
easyMDE.codemirror.on("change", () => {
    console.log(easyMDE.value());
});

Removing EasyMDE from text area

You can revert to the initial text area by calling the toTextArea method. Note that this clears up the autosave (if enabled) associated with it. The text area will retain any text from the destroyed EasyMDE instance.

const easyMDE = new EasyMDE();
// ...
easyMDE.toTextArea();
easyMDE = null;

If you need to remove registered event listeners (when the editor is not needed anymore), call easyMDE.cleanup().

Useful methods

The following self-explanatory methods may be of use while developing with EasyMDE.

const easyMDE = new EasyMDE();
easyMDE.isPreviewActive(); // returns boolean
easyMDE.isSideBySideActive(); // returns boolean
easyMDE.isFullscreenActive(); // returns boolean
easyMDE.clearAutosavedValue(); // no returned value

How it works

EasyMDE is a continuation of SimpleMDE.

SimpleMDE began as an improvement of lepture's Editor project, but has now taken on an identity of its own. It is bundled with CodeMirror and depends on Font Awesome.

CodeMirror is the backbone of the project and parses much of the Markdown syntax as it's being written. This allows us to add styles to the Markdown that's being written. Additionally, a toolbar and status bar have been added to the top and bottom, respectively. Previews are rendered by Marked using GitHub Flavored Markdown (GFM).

SimpleMDE fork

I originally made this fork to implement FontAwesome 5 compatibility into SimpleMDE. When that was done I submitted a pull request, which has not been accepted yet. This, and the project being inactive since May 2017, triggered me to make more changes and try to put new life into the project.

Changes include:

  • FontAwesome 5 compatibility
  • Guide button works when editor is in preview mode
  • Links are now https:// by default
  • Small styling changes
  • Support for Node 8 and beyond
  • Lots of refactored code
  • Links in preview will open in a new tab by default
  • TypeScript support

My intention is to continue development on this project, improving it and keeping it alive.

Hacking EasyMDE

You may want to edit this library to adapt its behavior to your needs. This can be done in some quick steps:

  1. Follow the prerequisites and installation instructions in the contribution guide;
  2. Do your changes;
  3. Run gulp command, which will generate files: dist/easymde.min.css and dist/easymde.min.js;
  4. Copy-paste those files to your code base, and you are done.

Contributing

Want to contribute to EasyMDE? Thank you! We have a contribution guide just for you!


Author: Ionaru
Source Code: https://github.com/Ionaru/easy-markdown-editor
License: MIT license

#react-native #react 

Marget D

Marget D

1618317562

Top Deep Learning Development Services | Hire Deep Learning Developer

View more: https://www.inexture.com/services/deep-learning-development/

We at Inexture, strategically work on every project we are associated with. We propose a robust set of AI, ML, and DL consulting services. Our virtuoso team of data scientists and developers meticulously work on every project and add a personalized touch to it. Because we keep our clientele aware of everything being done associated with their project so there’s a sense of transparency being maintained. Leverage our services for your next AI project for end-to-end optimum services.

#deep learning development #deep learning framework #deep learning expert #deep learning ai #deep learning services

Transfer Learning in Image Classification

The term Transfer Learning refers to the leverage of knowledge gained by a Neural Network trained on a certain (usually large) available dataset for solving new tasks for which few training examples are available, integrating the existing knowledge with the new one learned from the few examples of the task-specific dataset. Transfer Learning is thus commonly used, often together with other techniques such as Data Augmentation, in order to address the problem of lack of training data.

But, in practice, how much can Transfer Learning actually help, and how many training examples do we really need in order for it to be effective?

In this story, I try to answer these questions by applying two Transfer Learning techniques (e.g. Feature Extraction and Fine-Tuning) for addressing an Image Classification task, varying the number of examples on which the models are trained in order to see how the lack of data affects the effectiveness of the adopted approaches.


Experimental Case Study

The task chosen for experimenting Transfer Learning consists of the classification of flower images into 102 different categories. The choice of this task is mainly due to the easy availability of a flowers dataset, as well as to the domain of the problem, which is generic enough to be suitable for effectively applying Transfer Learning with neural networks pre-trained on the well-known ImageNet dataset.

The adopted dataset is the 102 Category Flower Dataset created by M. Nilsback and A. Zisserman [3], which is a collection of 8189 labelled flowers images belonging to 102 different classes. For each class, there are between 40 and 258 instances and all the dataset images have significant scale, pose and light variations. The detailed list of the 102 categories together with the respective number of instances is available here.

Figure 1: Examples of images extracted from the 102 Category Dataset.

In order to create training datasets of different sizes and evaluate how they affect the performance of the trained networks, the original set of flowers images is split into training, validation and test sets several times, each time adopting different split percentages. Specifically, three different training sets are created (that from now on will be referred to as the LargeMedium and Small training sets) using the percentages shown in the table below.

Table 1: number of examples and split percentages (referred to the complete unpartitioned flowers dataset) of the datasets used to perform the experiments.

All the splits are performed adopting stratified sampling, in order to avoid introducing sampling biases and ensuring in this way that all the obtained training, validation and test subsets are representative of the whole initial set of images.

Adopted strategies

The image classification task described above is addressed by adopting the two popular techniques that are commonly used when applying Transfer Learning with pre-trained CNNs, namely Feature Extraction and Fine-Tuning.

Feature Extraction

Feature Extraction basically consists of taking the convolutional base of a previously trained network, running the target data through it and training a new classifier on top of the output, as summarized in the figure below.

Figure 2: Feature Extraction applied to a convolutional neural network: the classifiers are swapped while the same convolutional base is kept. “Frozen” means that the weighs are not updated during training.

The classifier stacked on top of the convolutional base can either be a stack of fully-connected layers or just a single Global Pooling layer, both followed by Dense layer with softmax activation function. There is no specific rule regarding which kind of classifier should be adopted, but, as described by Lin et. al [2], using just a single Global Pooling layer generally leads to less overfitting since in this layer there are no parameters to optimize.

Consequently, since the training sets used in the experiments are relatively small, the chosen classifier only consists of a single Global Average Pooling layer which output is fed directly into a softmax activated layer that outputs the probabilities for each of the 102 flowers categories.

During the training, only the weights of the top classifiers are updated, while the weights of the convolutional base are “frozen” and thus kept unchanged.

In this way, the shallow classifier learns how to classify the flower images into the possible 102 categories from the off-the-shelf representations previously learned by the source model for its domain. If the source and the target domains are similar, then these representations are likely to be useful to the classifier and the transferred knowledge can thus bring an improvement to its performance once it is trained.

Fine-Tuning

Fine-Tuning can be seen as a further step than Feature Extraction that consists of selectively retraining some of the top layers of the convolutional base previously used for extracting features. In this way, the more abstract representations of the source model learned by its last layers are slightly adjusted to make them more relevant for the target problem.

This can be achieved by unfreezing some of the top layers of the convolutional base, keeping frozen all its other layers and jointly training the convolutional base with the same classifier previously used for Feature Extraction, as represented in the figure below.

Figure 3: Feature Extraction compared to Fine-Tuning.

It is important to point out that, according to F. Chollet, the top layers of a pre-trained convolutional base can be fine-tuned only if the classifier on top of it has already been previously trained. The reason is that if the classifier was not already trained, then its weights would be randomly initialized. As a consequence, the error signal propagating through the network during training would be too large and the unfrozen weights would be updated disrupting the abstract representations previously learned by the convolutional base.

#deep-learning #machine-learning #artificial-intelligence #image-classification #transfer-learning #deep learning

Mikel  Okuneva

Mikel Okuneva

1603735200

Top 10 Deep Learning Sessions To Look Forward To At DVDC 2020

The Deep Learning DevCon 2020, DLDC 2020, has exciting talks and sessions around the latest developments in the field of deep learning, that will not only be interesting for professionals of this field but also for the enthusiasts who are willing to make a career in the field of deep learning. The two-day conference scheduled for 29th and 30th October will host paper presentations, tech talks, workshops that will uncover some interesting developments as well as the latest research and advancement of this area. Further to this, with deep learning gaining massive traction, this conference will highlight some fascinating use cases across the world.

Here are ten interesting talks and sessions of DLDC 2020 that one should definitely attend:

Also Read: Why Deep Learning DevCon Comes At The Right Time


Adversarial Robustness in Deep Learning

By Dipanjan Sarkar

**About: **Adversarial Robustness in Deep Learning is a session presented by Dipanjan Sarkar, a Data Science Lead at Applied Materials, as well as a Google Developer Expert in Machine Learning. In this session, he will focus on the adversarial robustness in the field of deep learning, where he talks about its importance, different types of adversarial attacks, and will showcase some ways to train the neural networks with adversarial realisation. Considering abstract deep learning has brought us tremendous achievements in the fields of computer vision and natural language processing, this talk will be really interesting for people working in this area. With this session, the attendees will have a comprehensive understanding of adversarial perturbations in the field of deep learning and ways to deal with them with common recipes.

Read an interview with Dipanjan Sarkar.

Imbalance Handling with Combination of Deep Variational Autoencoder and NEATER

By Divye Singh

**About: **Imbalance Handling with Combination of Deep Variational Autoencoder and NEATER is a paper presentation by Divye Singh, who has a masters in technology degree in Mathematical Modeling and Simulation and has the interest to research in the field of artificial intelligence, learning-based systems, machine learning, etc. In this paper presentation, he will talk about the common problem of class imbalance in medical diagnosis and anomaly detection, and how the problem can be solved with a deep learning framework. The talk focuses on the paper, where he has proposed a synergistic over-sampling method generating informative synthetic minority class data by filtering the noise from the over-sampled examples. Further, he will also showcase the experimental results on several real-life imbalanced datasets to prove the effectiveness of the proposed method for binary classification problems.

Default Rate Prediction Models for Self-Employment in Korea using Ridge, Random Forest & Deep Neural Network

By Dongsuk Hong

About: This is a paper presentation given by Dongsuk Hong, who is a PhD in Computer Science, and works in the big data centre of Korea Credit Information Services. This talk will introduce the attendees with machine learning and deep learning models for predicting self-employment default rates using credit information. He will talk about the study, where the DNN model is implemented for two purposes — a sub-model for the selection of credit information variables; and works for cascading to the final model that predicts default rates. Hong’s main research area is data analysis of credit information, where she is particularly interested in evaluating the performance of prediction models based on machine learning and deep learning. This talk will be interesting for the deep learning practitioners who are willing to make a career in this field.


#opinions #attend dldc 2020 #deep learning #deep learning sessions #deep learning talks #dldc 2020 #top deep learning sessions at dldc 2020 #top deep learning talks at dldc 2020

Angela  Dickens

Angela Dickens

1597251300

Deep Learning Image Classification with Fastai

TL;DR

If you are beginning to feel burnt out on learning a subject it is often beneficial to take a step out of the weeds. I like to build something fun and easy to regain positive momentum on my learning journey. This is one project that helped get my creative juices flowing again. Load my notebook from my GitHub repository into Google Colab and upload the Kaggle data set to learn how to build an image classifier using the fastai software! Don’t forget to set the hardware accelerator to GPU!

P.S. Directly upload notebook to Colab by going to File →Upload Notebook → GitHub Tab→BSamaha/Chest-Xray-FastAI

Image for post

Image for post

How to easily upload a GitHub Notebook to Google Colab


Intro

The Dunning-Kruger Effect

You have set a goal such as wanting to be a data scientist or data engineer. Your passion and energy are through the roof as you rifle through all the material on the subject you can find. Perhaps you even started a coding boot camp or a Coursera class to be your guide on your path to data mastery. Your brain becomes saturated with dopamine as you quickly rack up quick wins. First, it was basic python programming. Then you made your first linear regression with sci-kit learn — machine learning isn’t so bad after all! You quickly absorbed all information thrown at you. Nothing can stop you.

I’ve been there, and I still go there. I am human after all. I’m referring to the peak of “Mt. Stupid” in The Dunning-Kruger Effect. Linear regression was easy with a perfectly curated data set, but that’s not reality. Finally, after about a month or two of flying through the material with ease, you hit a wall named statistics. You grind through it and barely make it out — except you feel like you haven’t retained anything. Do people remember all this distribution stuff? Spoiler Alert: yes, and you will be questioned about them. As you pursue on you realize your python skills are nowhere near where they should be. You decide to look at a couple of job postings and then it hits you. You don’t have an advanced graduate degree in computer science, and you don’t know how to pronounce some requirements of the job much less know what they are — Hadoop? Welcome to the Valley of Despair. You want to quit and move on with your life.

#deep-learning #machine-learning #deep learning