The International Image Interoperability Framework (IIIF) Audio/Visual (A/V) Technical Specification Group aims to extend to A/V the benefits of interoperability and the growing ecosystem of clients and servers that IIIF provides for images. This repository contains user stories and mockups for interoperable A/V content – contributions are welcome.
Home Page: http://iiif.io/community/groups/av/
License: Apache License 2.0
Use a SC Canvas extended with time to associate the resources with. The time dimension on the Canvas provides the virtual timeline.
Source: BL workshop notes
Interest: 100%
Custom media players for file types that aren't directly supported, or subhosted via youtube. Something distinct from just a video file, but actually an iframe.
Source: BL workshop notes
Interest: 90%
Issue: postMessage() to communicate with iFrame; maybe standardize interface? Would be super valuable. Look at youtube API. Likely need a way to describe the interface supported. This may need a separate spec; see quick survey of state of art.
Initial implementations could probably hardcode use of a few site-specific APIs such as YouTube’s to demonstrate basic effectiveness.
[If folks are interested in further followup on iframe embedding standardization, see https://github.com/iframe-player/iframe-api-docs where @brion will be collecting details and trying to organize something. Feel free to add ideas there as issues. Will try to make sure editable pages are up or can add people.]
Use of profile for the API to use, e.g. for youtube?
Multiple remote services, like multiple formats. Toss in a YouTube link, and a local MP4, and let the player decide which to load based on availability.
Scraping YouTube for the raw mp4 would be so much easier in theory… but unreliable, and won’t work with content protection, etc. Would also violate their ToS.
Stability of iframe URLs, parameters? (Seem to mostly be stable for YouTube and Vimeo.)
Extra requirements and oddities -- youtube says “must” be 200x200px (see https://developers.google.com/youtube/iframe_api_reference) What’s that mean? :D
Comparing with http://wellcomelibrary.org/iiif/b16748967/manifest example:
"rendering": [
{
"@id": "https://s3-eu-west-1.amazonaws.com/wdl-video-open/mp4/80e62970-d26a-403e-8aa8-480de6d22495.mp4",
"format": "video/mp4"
},
{
"@id": "https://s3-eu-west-1.amazonaws.com/wdl-video-open/webm/80e62970-d26a-403e-8aa8-480de6d22495.webm",
"format": "video/webm"
}
],
Let’s imagine extending that to also include a remote link and an API reference, something like:
{
"@id": "https://www.youtube.com/embed/dQw4w9WgXcQ?enablejsapi=1",
"format": "text/html",
"profile": "https://developers.google.com/youtube/iframe_api_reference",
}
Or more simply with a fake type, where the player would know how to add the api-specific iframe parameters:
{
"@id": "https://www.youtube.com/embed/dQw4w9WgXcQ",
"format": "x-video/youtube"
}We might want to look to the service pattern to embed the YouTube @id & profile as above (see image annotation resources in the Presentation API). [This could avoid needing a separate info.json for every youtube video you want to embed?]
Note use case given where YouTube video was used preferably but a local file needed to be provided for China. Providing both the embed link and the .mp4, in preferred order, should work same as including an .mp4 and a .webm: player can know that YouTube won’t work locally and skip over it.
John Dyer of MediaElement.js was working on an HTML5 Video Wrapper for the YouTube and Vimeo embedded player APIs (seems to still be maintained as part of MediaElement.js). This is very similar data model, taking the embed URL in a
The shim itself: https://github.com/johndyer/mediaelement/blob/master/src/js/me-shim.js
Popcorn.js has something similar: http://popcornjs.org/popcorn-docs/media-wrappers/
Client feature: Display of audio as waveform or sonogram. For example, speech and music can be easily seen in waveform. Sonogram for wildlife. [Generalize along with audio postprocessing?] Web audio APIs let you generate this. Way to refer to a waveform file that was pregenerated? https://developer.mozilla.org/en-US/docs/Web/API/Web_Audio_API
client feature only? Question as to how to implement it efficiently. Could have special service / seeAlso block that delivers the data for the waveform, rather than calculating it from the audio stream. Can something be identified and adopted for this?
Soundcloud API for waveforms: https://developers.soundcloud.com/blog/waveforms-let-s-talk-about-them
http://www.waveformjs.org
http://www.bbc.co.uk/rd/blog/2013/10/audio-waveforms
https://github.com/bbcrd/waveform-data.js/blob/master/README.md
Source: BL workshop notes
Interest: 50%
Content protection. View the content but don't need to authenticate. Mostly for IPR issues. HTML5 video player, solution was to use a unique token every time. Play content once with the token. Can go forwards and backwards. But on a download manager you can't get it. Stronger than Image API protection. Would need to webscrape javascript.
Need to pass token to the streaming environment. Could be in URL params, cookie, header etc.
Tags: download
IIIF: Auth API
Issue: Question as to whether a redirecting video/audio element src redirects might cause problems? Otherwise can have a IIIF implementation of the Auth API, either as a shim on top of existing system or as a new system.
Auth API looks like it should work with HTML5 video/audio as long as resource is on the same domain as auth server (cookies), but still allows downloading the local file from the browser once authenticated.
Solution: Specify token service at level 3?
MPEG-DASH and HLS can make it hard to download.
IIIF Tile Harvester in Python:
https://github.com/azaroth42/iiif-harvester/blob/master/tilescraper.py
Source: BL workshop notes
Interest: 60%
Audio manifest only. No x,y at all, just duration.
Need to make x,y dims for canvas optional (iff t given)
Source: BL workshop notes
Interest: 100%
Non-linear time alignment. Different tempos of recording of music. Align points, but not in time. Same score, different audio times. (http://beethovens9thsymphony.touchpress.com/ is a realisation of this use case). See also NYPL Labs video web tool to combine multiple recordings (http://digitalcollections.nypl.org/tools/video/)
Wikipedia use cases. Music / speeches where would want to mark up points from one alternative version to another.
Might be able to be done automatically. Good to look at NYPL offering.
Refer to multiple points at the same time, and be able to state they're the same abstract thing.
Annotations on video, with multi-camera. Diff positions based on version of video being watched
Annotation that tags multiple canvases to indicate aligned points in the different canvases.
Source: BL workshop notes
Interest: 65%
Video API @id, can reference info.json for rights. In Presentation API, can have link to service.
Source: BL workshop notes
Interest: 60%
Level 4! :) Put name of track into the media API slot. Very high in the fruit tree.
Source: BL workshop notes
Interest: 60%
Ability to describe a master file that's not available. Might have only delivery of different formats from it. Do publish metadata about it. Want consistent time codes between master and delivery copy. Behind the scenes implementation detail that user doesn't care about. e.g. someone annotated interesting part of derivative video, can then go back to master to re-extract more high quality version.
Scope: ?
Implementation detail? System needs to know the background mappings (e.g. jp2 -> png in Image API, without making jp2 available).
Give access to full quality, but only small segments. Max but for temporal region, rather than size.
Describe the image qualities of a single frame extracted from the underlying video source.
Turn video/canvas+time into image api identifier, and fetch info.json for it. -> relates to Download still frame from video as image use case and how that service is described
Source: BL workshop notes
Interest: 60%
Slideshow view of images with audio. Paint the canvas over time with image annotations.
E.g. Audio podcast about painting, bring up image of painting. Audio with slideshow.
[Related to missing scene] woord.nl
One canvas with images painted for particular time ranges (existing case of no time range means always present over whole duration), alternate solution is to have a sequence with one image each and a segment of the audio.
Related IIIF: Annotation, Canvas
Source: BL workshop notes
Interest: 100%
(Geo)tagging of content. e.g. link to streetview of the scenes. Movie locations. Or tagging people with identities such as ancestry or orcid. [Related to tagging]
Annotations, motivation: tagging.
Source: BL workshop notes
Interest: 70%
Ability to control playback. Should be able to play, pause, skip ahead/back, change volume, adjust balance (over channels).
For case of HTML5 all of these facilities are supported. Events will be different for iframe scenario. LevelA? LevelImpossible? For now be able to say “this iframe supports youtube/whatever api” for extended clients to support if they want? In the future try to standardize it, but will have to get Google on board probably (eek!). [NB I think it would be extremely useful to have a spec for this that academic-based media providers could implement, even if YouTube is not on board. -Jon D.]
Interest: 100%
Source BL workshop notes
Describe media type, aspect ratio and resolution. e.g. if it's 4/3 or 16/9 aspect ratio
Properties on info.json.
Source: BL workshop notes
Interest: 100%
Useful to surface the time code that's currently played. Want to trigger events based on time point being played in the browser. HTML5 video can be queried for time points. Popcorn. Offset from start point? Multiple items playing (need global clock)? Custom iframe players (youtube assets in iframe, etc) - how does player get the low-level time?
Easy with regular video stream, as get it from HTML5. Hard with iframe, needs spec.
Source: BL workshop notes
Interest: 100%
Internationalization: Audio streams in welsh / english, along with subtitles in Welsh/English.
Embedded textual content handled via JSON-LD. Options:
For situations when there are more than one video associated with a canvas, it might be useful to associate the subtitles (etc) with the canvas as individual annotations.
For situations when there is a single media stream, it would be easier to use existing subtitle resources and have the player render them. → Reference from info.json to different language subtitle files (outside of presentation API)
Subtitles that should be rendered by the viewing application in a separate window/panel should be annotations on the canvas rather than a subtitle file.
Transcription (with speaker identification) != captioning
SRT “spec” http://srt-subtitles.com/ & https://en.wikipedia.org/wiki/SubRip - very simple entries: entry number, time range, text (with limited HTML)
Use case for zones: associate timestamps of subtitles with the zone, and then associate zone with the time frame of multiple canvases.
Can associate annotations with the canvas.
Link from video-info.json to a zone.
Source: BL workshop notes
Interest: 100%
Get access to a full video/audio stream and play it.
Media API equivalent to Image API:
Level0 equivalent is standalone content where either the server doesn’t support byte ranges (!!) or the format makes this useless anyway (one has to download file)
Level1 as more dynamic. Allow range-based seeking within file. (.mp4 or .webm on apache/nginx/etc, or HLS or MPEG-DASH) - expected to be the common baseline
Level2 provide downloads of rewritten subsets of media content on demand.
LevelA use of existing APIs (e.g. youtube/vimeo/soundcloud) -- need different information in the service description for how to interact with the API. Might a different service? info.json equivalent could be also translation table to other existing APIs. (iframe API issue survey)
NB Can seek within a video stream from the client more easily than "seeking" into image. If need to play from A to B in video, only have to load the header and then seek to point A via HTTP ‘Range’.
Note also that some file types don't support seeking. Need to expose seekable/unseekable as metadata (supports: …)?
Source BL workshop
Playback (audio and video) at different speeds. Multiplier on the speed. Open question as to whether its client or server.
Issue: If there are multiple video/audio on a single canvas at the same time, then the multiplier applies to the canvas not the stream, otherwise the time point of the canvas is wrong.
Whether to resample audio or not. (control in scope?)
Client application concern, not interop.
Source: BL workshop notes
Interest: 100%
Viewer level: user level comparison of video (two manifests/sequences or two canvases). (2-up) [does this need synchronized playback between both instances? maybe?]
client feature simple matter of lines of code :)
Source: BL workshop notes
Interest: 75%
[See reference use case]
client feature, to send URI of range / track to the playlist thing
Source: BL workshop notes
Interest: 50%
Full text search of transcribed speech. from speech to text, rather than OCR or manual.
Search API for annotations. If it's not annotations, then it's out of scope
Source: BL workshop notes
Interest: 100%
Annotate (refer to) the audio and video channels separately, even if displayed with both at once and part of a single representation (video stream with audio). Example of annotating the instruments vs technique vs music against different streams. Layers of annotations.
Use media fragment #track= in the URI and then list the available track names in the video.json. Default name of "video".
"tracks": [
{"value": "audio", "language": "en", "type": "Audio", "label": {"en": "The English Audio Feed"},
{"value": "subtitles", ...},
{"value": "video", ...}
]
Issue: Hard! :)
Source: BL workshop notes
Interest: 80%
Be able to describe navigation and hierarchy. Tracks /song within album regardless of carrier. Logical structure.
Related IIIF: metadata on Range
Source: BL workshop notes
Interest: 100%
Challenge: Twitter needs content to come from a whitelisted host. (A particular presentation player, such as Universal Player, could expose the ‘player card’ interface and be whitelisted, which should allow any AV resource to be playable as long as linked through there.)
https://dev.twitter.com/cards/types/player
Facebook Open Graph:
https://developers.facebook.com/docs/sharing/webmasters#media
https://www.jwplayer.com/blog/publish-your-videos-to-facebook-with-a-jw-player/
Issue: Tricky to get it to work. Consider also sharing segments of video (e.g. segmentation in media api). Support for cards at the API level okay, but need to be approved.
Not clear that there is anything needed for specifications here.
Source: BL workshop notes
Interest: 65%
As a lay user, I want to control when the audio stream starts and stops at the individual page level, rather than line by line or note by note, in order to not have to click many times and to experience the full page worth of music.
(do you know of, or can you imagine, similar use cases?)
It should be possible to progress from one audio annotation to the next on a single canvas, in the same as progressing from canvas to canvas with or without a gap as per #12. Thus auto-play annotation two after playing annotation one.
Annotation List could provide a hint as to the continuous-ness of the annotations. Lists are ordered already, so it would be a case of playing from the beginning of the list.
If there are multiple points to break at, this could be multiple lists each with the hint, or the connection could be at the annotation level rather than the list level.
From Shared Canvas demo2, Worlde's Blisce.
Related IIIF: Sequence (with some other viewingHint, and maybe some other prop. for length)
If the controller should reset the time to zero, then you need a new canvas in a sequence. If the controller should continue in the same time then it should be a zone on a single canvas.
Canvas for the gap with short duration. viewingHint to say "gap".
For single canvas case, might either have a zone or just leave the time space empty.
(do you know of, or can you imagine, similar use cases?)
(any ideas about how your use case might be supported)
Source: BL workshop notes
Interest: 100%
Download still frame from video as image. Or series of frames as images.
Level 3 (unless implemented in client). Output is a different type (image) which distinguishes this use case from the selection of a segment of video. Media fragment spec provides way to specify point in time / frame to extract.
For bonus points: create Image API service for the frame
put video identifier + region/frame as identifier of image API ?
Source: BL workshop notes
The use cases should have diagrams where possible to help explain the issue and proposed solution.
Valuable to do together at The Hague, if not before if time allows.
Diagrams to be managed in the repository, and linked from the individual issues. This will let us build a use case document as we progress from the issue descriptions and images.
Refer to a spatial bounding box of video (by area)
xywh media fragments (same as for images)
Related IIIF: Canvas # Fragments
Source: BL workshop notes
Interest: 100%
Implication: Assume default behavior is user actuation of control to step from canvas to canvas. Use continuous hint to change behavior. Note that need to queue up next canvases resources before finishing the first one (and audio processing issues to get nice continuation of one track to another). [Sample-accurate gapless playback is possible via Web Audio. There are libraries existing to help with this which a player can use.]
Note alternate option of putting multiple tracks on one canvas. Here the amount of overlap, gaplessness, or gaps would be determined by positions on the shared canvas.
Related IIIF: Sequence w/ e.g. viewingHint=”continuous”
Source: BL workshop notes
Interest: 100%
Access restrictions: prevent replay of part of a recording, e.g. redacting. By time code. On site vs external access. Geo-locking for out of copyright in UK but still in for US.
Annotation of the restriction
pre-constructed versions, with redirects. Not a client side thing, as clients could simply ignore it. So can't expect client to not play something it has access to.
IIIF: Auth
Source: BL workshop notes
Interest: 90%
Missing or lost scene. Virtual reconstruction. Blank time space that can be annotated.
Source: BL workshop notes
Interest: 40%
Increasing / decreasing audio delay. Same with subtitles.
Could be variation of synchronization use cases.
Q: Requires subframe accuracy?
Seems like a client feature to provide a control, or not.
Source: BL workshop notes
Interest: 50%
As it's expensive to create snippets (either by time, region or both) it would be useful to request a snippet, and then be notified later that the content is available for download. This would only be for download, rather than for playback.
From #4 via @glenrobson
Client should be able to request an arbitrary segment of a resource by time for download.
This is distinct from:
Allow temporal dimensions by importing the media fragment spec (SMPTE) into the Media API.
See: https://www.w3.org/TR/media-frags/#naming-time
Thus the URL pattern might be something like:
identifier/t1,t2/xywh/size/?/default.avi
It has implications on an info.json equivalent, where pre-configured temporal "tiles" might be available for download, either by track/chapter/semantic section or simply by subdividing the content evenly.
Difficult to do on the fly, so likely a "level 3" feature
Source: BL workshop notes
Thumbnails of audio/video -- image that represents the video, then sub-thumbnails for time based ranges. For video, and for ranges / different shots.
Use existing mechanisms for association of thumbnails with Manifest, Sequence, Canvas, Range etc. Question of how to deal with more structured sets of thumbnails for key-frames etc - use annotation list of images, might need paging for many images (e.g. every 5s for a few hours).
Source: BL workshop notes
Interest: 100%
rights fields from image/prezi api
Source: BL workshop notes
Interest: 100%
Based on the discussion at the Hague working group meeting, we discussed some potential solutions for requesting transformations to videos that mirrored the kinds of transformations that can be done with the Image API. The set of possible transformations for video (and even audio) is potentially very large. Before we begin trying to specify what an A/V bitstream API might look like and because there is a large number of videos currently delivered, it would be good to get a better sense of what is most important. We can begin experimenting with transformations and possible parameters before trying to specify a bitstream API by giving information to clients through the Presentation API. One proposed solution which appeared to be a productive direction is the use of Selectors for manipulations on a video or audio file.
We could use a mock-up of what selectors might look like for video and audio. The example(s) would be best if they're within the context of a full manifest.
Image API Selectors:
http://iiif.io/api/annex/openannotation/#iiif-image-api-selector
Draw on the video. - do it yourself football punditry
Paint svg / image on to a timed segment of the video/canvas
Source: BL workshop notes
Viewer facilities needed to meet accessibility requirements. Eg, annotate descriptive audio tracks [Talk to DAISY about audio books]
Viewing Hints (and other metadata) to make things more accessible?
See also WAI-ARIA https://www.w3.org/WAI/intro/aria .
? https://www.w3.org/WAI/WCAG20/quickref/#media-equiv ?
Ability to turn on/off multiple audio/video tracks for visual/auditory impaired (e.g. for Descriptive Video Service audio track).
Annotations with oa:describing motivation.
Source: BL workshop notes
Interest: 100%
Be able to request an arbitrary region of the entire video by xywh, rather than by time. This could be used for:
This is distinct from:
Implication is for download, and that the server will construct a full stand alone video representation.
Incorporate region into the AV API, in the same way as for the Image API, using the Media Fragment syntax.
Has implications on an info.json equivalent for the AV API where different "tiles" of the video could be pre-generated up front. Would need to be distinct from the entire video at different resolutions.
Probably a "level 3" feature due to complexity.
Source: BL workshop notes
Interest: 100%
Zoom into videos. CSI "enhance" feature or Bladerunner :) Perhaps presentation issue. Load high resolution version of video, rather than cropping and tiling of the video.
Source: BL workshop notes
Interest: 100%
Scope: Version >1
Set of content that includes audio (mp3) plus JPG plus PDF etc. Might be scanned physical carriers. [related to show the cover etc]
Collection of manifests, either with or without multi-part. Or a manifest with sequence of timed / non-timed canvases
Source: BL workshop notes
Interest: 100%
Ability to refer to a point in time. Standard syntax for addressing points. Want to say this chord at this point. SMPTE time codes. Might also need sample level access on audio files -- this 10 sample window. Frequency content of the window. SMPTE, Samples and Time. HTML5 video not great at supporting sample accuracy, get nearest key frame based on time. Could annotate with sample accuracy, but just works as metadata.
Related IIIF: Image API Region, info.json, # Fragment on Canvases
Media fragments support NPT, SMPTE (https://www.w3.org/TR/media-frags/#naming-time) and note no problem combining t=123.45&xywh=1,2,3,4 (https://www.w3.org/TR/media-frags/#processing-name-value-lists )
Media fragment spec does not specify mandatory or canonical order of parameters. If using in IIIF we’d probably want to mandate (or at least recommend a canonical order) to avoid the creation of duplicate URIs for the same thing.
Film folks need frame specific references. Can build a more advanced player that supports it. Frames would need to be relative to original scan of the film. Transcoding gets … "exciting".
e.g. freeze on a frame where the original physical film has a scratch. Instead of referring to a point in film by time, we can do it by frame as well
SMPTE allows reference to a frame in the format: hh:mm:ss:ff
eg: 00:01:30:02 is the third frame in 1 minute 30 seconds. (SMPTE spec is closed, but https://en.wikipedia.org/wiki/SMPTE_timecode describes)
Source: BL workshop notes
Interest: 100%
Control of multi-channel audio, such as spatial information as to where the audio should come from. BL have a considerable number of ambisonic recordings (https://en.wikipedia.org/wiki/Ambisonics )
Media-fragment spec has track fragment, no notion of channel. Could track be sanely used as channel? There appears to be explicit alignment between tracks in the webaudio spec and tracks in media-fragment spec so it would seem wrong to try to make a map to channels.
Level 3 server could separate them into individual streams.
Web Audio API supports arbitrary number of channels already. Could redirect/remix as necessary.
Might be able to work from video element (?)
http://webaudio.github.io/web-audio-api/#MediaElementAudioSourceNode
Requires suitable CORS headers on the source video
HTML5: https://www.w3.org/TR/html5/embedded-content-0.html#media-resources-with-multiple-media-tracks
Workaround is to encode channels as separate tracks and list them as such.
Source: BL workshop notes
Interest: 50%
We would like to see some proposals for what folks think an info.json for A/V might look like so that we can make further decisions on what we need in an information package about a video or audio recording.
Yes! Propose solutions. 😄
Notes from the Hague meeting that can help to inform potentially productive directions.
https://docs.google.com/document/d/1cnkOPm7rC9uKeSxorFpu004ZzJxolVLWex5NdxMKnHY/edit
Scope: Auth API
Issue: some video servers require putting a token in the URL, rather than a cookie.
Source: BL workshop notes
Interest: 100%
Requirements will vary depending on whether the subtitles are stored with a video/audio or whether they are separate (e.g. crowdsourced)
Related IIIF: Annotations (Layers and List), Canvas # Fragments
Source: BL workshop notes
Interest: 100%
Play multiple (synchronized) videos. Multi-camera and mic recordings in parallel. For example: Use of multi-camera work for analysis of user behaviour. Or capture of slides and audio and video for a lecture.
Easy case of multiple video choice on a single canvas is easy (Choice). Harder case with multiple synchronized devices is the same as the previous use case.
Multiple: multi-screen presentation. Big screen for video, and separate screen for separate videos to be synchronized. Museum exhibits. e.g. world's fair. http://www.eamesoffice.com/the-work/think/ is the uber-AV use case - can we reproduce this? (there’s a 3D aspect to this too).
Difference: Presentation of the content. Multi camera might have a timecode to sync across different streams. Presentation mode should (hopefully) be consistent.
Requirement: Do not assume that all content is rendered by the same device.
(any ideas about how your use case might be supported)
Source: BL workshop notes
Interest: 60%
Access via Mobile (responsive), or embedded on a page. Fully featured vs slimline player. Note that Apple require HLS for video over 10 minutes in iOS apps. [Not universally enforced but it’s always a concern.]
Yes :) Different viewing applications can be built.
Source: BL workshop notes
Interest: 100%
Need to be able to control playback (volume, time, etc) of non AV-API content, such as Youtube.
Add a service block to an AV resource in the Annotation that associates it with the canvas:
{
"type": "Annotation",
"target": "http://example.org/canvas/1",
"body": {
"id": "http://youtube.com/video",
"service": {
"id": "http://youtube.com/video/api",
"profile": "uri-for-youtube-api",
"more": ["params", "go","here", "too"]
}
}
}
Expands on #2
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
Django
The Web framework for perfectionists with deadlines.
Laravel
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft
Open source projects and samples from Microsoft.
Google
Google ❤️ Open Source for everyone.
Alibaba
Alibaba Open Source for everyone
D3
Data-Driven Documents codes.
Tencent
China tencent open source team.