Video Indexer
Welcome to the Video Indexer Forum
Categories
API – Any ideas or feedback pertaining to features or enhancements to Video Indexer API.
General – Any general ideas or suggestions related to Video Indexer.
Portal – Any ideas or feedback regarding the web portal for Video Indexer
Video AI – Any ideas or suggestions regarding Video AI features in Video Indexer.
-
Landmark Recognition For Videos
Looking for the ability to recognise landmarks (e.g. Eiffel Towel, Sydney Harbour Bridge, etc..) present in videos that are uploaded and stored within Video Indexer.
I believe this capability exists within the Computer Vision API today (9,000 landmarks when I last checked)
8 votes -
Import existing edited transcripts or caption files
The ability to upload an already edited transcript will enable those with existing libraries to move over content without the need for further human editing.
Ideally VI would apply the timings, and insert them in their appropriate place on the Transcript panel.
3 votes -
Create an Azure Media Services 'Media Processor' as a wrapper for Video Indexer.
Media Processors in Azure Media Services are the components you can string together in a pipeline to build a media workflow, including transcoding and media analytics, etc. In order for Video Indexer to more cleanly integrate with existing media workflows, there should be a VI Media Processor that can be added to regular Azure Media Services jobs. This way, I can create an Azure Media Services job in a few lines of code that will upload a video, transcode a mezzanine format such as ProRes into a multi-bitrate mp4, generate a shortened version for a video thumbnail using the video…
5 votesMany of the insights provided by VI are available in AMS RPv3 audio and video analysis.
VI also allows providing an AMS asset as an input for indexing -
Use web hooks for indexing job completion notification
When I submit a video to be indexed via UploadVideo API method, I'd like to provide a web hook url as a callback mechanism to indicate when a video is done processing. Processing state could also be provided via this web hook mechanism as well, not just to indicate completion.
The advantage to this approach is a decoupling of the submission of indexing jobs from the post completion handling. This is important when using Azure Functions since you don't want your functions to be running (polling for progress) for as long as it takes the video indexing job to complete.…8 votesThis is available through the callbackUrl parameter you can send in the upload video request – https://api-portal.videoindexer.ai/docs/services/Operations/operations/Upload-Video
-
Disable quarantine for strictly private video
Videos that are used for preparation of subtitles for commercial movies/series are sometimes quarantined by the VI (i.e. Dr.House). The VI could be used to provide a transcript of the video which will be manually corrected/adapted to a high quality subtitle file. However in this scenario the VI shall not put commercial movies/series into quarantine because people preparing subtitles always get video material even before it is shown in theaters or on TV and they would like to send it to the VI to get the transcript.
One way to avoid putting videos into quarantine is to introduce "Strictly private"…
2 votes -
Provide an API function that returns the list of supported languages
It would be useful to get the list of supported languages to be able to display available languages in the UI of the application using the VI.
2 votes
ohadjas
responded
Can you please clarify – list of supported languages for transcript or translation?
-
Allow setting language to "None" to skip speech recognition for unsupported languages
If the audio language is not supported by the VI then the breakdown might still be useful but speech recognition does not make any sense in such case. Setting the language to "None" or "Other" or "Unsupported" would be useful.
3 votes -
Add edit ability for transcripts to correct auto generated transcripts
Inevitably there are going to be words that are misidentified by speakers and it would be great if there was the ability to correct the speech to text manually. Perhaps this could be used to feed back into the model for training as well.
This would make the transcript feature much more useful as basically it could function as an assisted transcription tool where all an editor would have to do would be to supervise the generated transcription correct the parts that were not correct and save.
2 votes -
Allow tagging of faces that are not detected to existing entities
In the case that a celebrity is not detected by the AI but is in fact in the video, it would be great to allow the tagging to an existing entity.
7 votes -
Ability to re-run Cognitive process
The backend Cognitive Services is continuously improving and the results will likely be improved when the same video will be uploaded again.
Re-running cognitive processing to already uploaded videos will meet the situation.4 votes -
Ability to train face model for people not matched by celebrity database
The API should expose a way to train a custom face model with provided identification for faces that were not matched against the celebrity database. Additionally, there should be a way to correlate new recognized face ID with previous instances of the unknown face (in customer db).
An example API call would include a faceId (retrieved from breakdown), a name (caller provided identification), and an optional CustomId field (for application tracking, which would subsequently be returned in future breakdowns).
An additional API call that would be useful would allow a batch training of the model, even before any breakdowns are…
16 votes -
Save button
The Save button is under a hamburger menu !?
Screenshot: http://imgur.com/a/pwjjpThis is bad for discoverability. Takes a few seconds to figure out.
1 voteThis is live now in the VI Portal! Thanks for your suggestion, it is very appreciated.
Keep the ideas coming. -
Identify speakers in audio only
Idetify speakers with audio only, not faces, this is useful for Radio content
3 votes -
Add scene detection
Add scene detection to allow viewers to jump to different scenes in a video.
2 votes -
Add Logo detection
Such as, Looking for the number of "Coca Cola" logos in my video. Either search for well-known famous images or search for matches to a given set of uploaded custom image files.
15 votesThank you for you feedback, this request is in review. We will update with our plans for logo detection.
-
Add "Shazam-like" detection of well-known/copyrighted music
Recognizing famous voices, sound effects, or even decibel level measurement (ie. how loud crowd noise is) would be great too
7 votes -
Find the black frames within a video
Provide both the total count, timecode of each, and total length (MM:SS)of black frames in the video, both with and without silence
2 votesBlack frames detection is provided in the annotations section
-
Import into VI all of the videos within a particular folder/container
Rather than a single file at a time, add the ability in the UX or API to upload any of the following: 1) an entire local folder's worth of videos 2) , a Container within Azure Blob Storage 3) Dropbox folder, or 4) AWS S3 folder
4 votes -
Add live ingest
Ability to capture a live stream and analyse it in real time.
16 votes -
Undo remove person
I just removed a person, and I can't figure out how to put them back.
4 votes
- Don't see your idea?