jina-ai / executors

internal-only
Apache License 2.0
31 stars 12 forks source link

Encode/decode for different executors #240

Closed CatStark closed 3 years ago

CatStark commented 3 years ago
Done Executor Name repository Reference Asignee
Mp3Loader executor-audio-mp3loader https://github.com/jina-ai/examples/blob/master/audio-to-audio-search/executors.py#L79:L97 @jakob1111996
WavLoader executor-audio-wavloader https://github.com/jina-ai/examples/blob/master/audio-to-audio-search/executors.py#L79:L97 @jakob1111996
VideoFrameLoader executor-video-frameextractor, executor-video-loader https://github.com/jina-ai/eap-google-ads/blob/main/pods/executors.py#L47:L112 @makram93 & @nan-wang
VideoAudioLoader executor-audio-frameloader https://github.com/jina-ai/eap-google-ads/blob/main/pods/executors.py#L47:L112 @makram93
VideoSubtitleLoader executor-subtitle-frameloader https://github.com/jina-ai/eap-google-ads/blob/main/pods/executors.py#L47:L112 @makram93
PptxLoader executor-pptxloader https://support.microsoft.com/en-us/office/extract-files-or-objects-from-a-powerpoint-file-85511e6f-9e76-41ad-8424-eab8a5bbc517 @cristianmtr

Reference

jakobkruse1 commented 3 years ago

Just to understand this correctly: Do we want one executor that can read every data type or should we have separate executors for separate data types? @CatStark @nan-wang

nan-wang commented 3 years ago

Just to understand this correctly: Do we want one executor that can read every data type or should we have separate executors for separate data types? @CatStark @nan-wang

@jacobowitz sorry, I read the message wrongly. For loading each type, We want to have one executor.

maximilianwerk commented 3 years ago

The main point here is, that it is not just a data loader from somewhere but at the same time a decoder. E.g. the Mp3Decoder should:

maximilianwerk commented 3 years ago

Decision from alignment meeting: We should have one Decoder/Loader per media type for now. One AudioLoader and one VideoLoader.

jakobkruse1 commented 3 years ago

Just to understand this correctly: Do we want one executor that can read every data type or should we have separate executors for separate data types? @CatStark @nan-wang

@jacobowitz sorry, I read the message wrongly. For loading each type, We want to have one executor.

Alright, I understand :) Btw, I am @jakob1111996, not @jacobowitz (This would be Tobias)

makram93 commented 3 years ago

Decision from alignment meeting: We should have one Decoder/Loader per media type for now. One AudioLoader and one VideoLoader.

Hi @jakob1111996, in yesterday's meeting we decided to have one executor per mime type. So, overall three executors - audio, video, and pptx. I will be working on this ticket too and am picking up video executors if you haven't started with video already. Just wanted to let you know so that we both don't end up doing the same work. :)

jakobkruse1 commented 3 years ago

Nice @makram93 , I have started working on Audio, so we should be fine :)

makram93 commented 3 years ago

@nan-wang, after discussion with @florian-hoenicke, we decided to skip the subtitle encoder for now as we don't see any use case currently.

jakobkruse1 commented 3 years ago

For the audio reader, which data types should be supported? Right now, I am supporting MP3 and WAV. Any other types you think are necessary? AIFF, WEBM, FLAC, AAC, WMA, MP4, M4A? @nan-wang @makram93 @maximilianwerk

nan-wang commented 3 years ago

For the audio reader, which data types should be supported? Right now, I am supporting MP3 and WAV. Any other types you think are necessary? I suggest not supporting other types at the moment. Let's see what are requested by the community re after launching the hub.

nan-wang commented 3 years ago

after discussion with @florian-hoenicke, we decided to skip the subtitle encoder for now as we don't see any use case currently We are actually using this in one of the EAPs although this part is currently implemented by the user.

florian-hoenicke commented 3 years ago

We are actually using this in one of the EAPs although this part is currently implemented by the user.

@nan-wang okay, I see. What video source files should be supported for subtitles?

nan-wang commented 3 years ago

okay, I see. What video source files should be supported for subtitles Let's focus on .mp4 and save the outputs at .srt using ffmpeg.