14:26:57 RRSAgent has joined #webmachinelearning 14:26:57 logging to https://www.w3.org/2018/12/13-webmachinelearning-irc 14:27:10 Zakim has joined #webmachinelearning 14:27:24 RRSAgent, make logs public 14:27:40 Meeting: WebML CG Teleconference – 13 December 2018 14:27:42 Chair: Anssi 14:28:04 tomoyuki has joined #webmachinelearning 14:28:10 Scribe: Anssi 14:28:14 Agenda: https://github.com/webmachinelearning/meetings/blob/master/telcons/2018-12-13-agenda.md 14:28:31 Present+ Anssi_Kostiainen 14:28:38 Present+ Ningxin_Hu 14:28:42 Present+ Tomoyuki Shimizu 14:28:46 Present+ Greg_Whitworth 14:29:20 Present+ Chris_Heilmann 14:29:29 scribeNick: anssik 14:29:33 RRSAgent, draft minutes v2 14:29:33 I have made the request to generate https://www.w3.org/2018/12/13-webmachinelearning-minutes.html anssik 14:30:09 TOPIC: Welcome 14:30:51 Rafael has joined #webmachinelearning 14:31:11 Yuan has joined #webmachinelearning 14:31:13 anssik: welcome to the first WebML CG telcon. We'll meet on a monthly cadence when there's an agenda. 14:31:29 n+ Rafael 14:31:46 Present+ Raphael_Cinton 14:32:02 anssik: subscribe to https://github.com/webmachinelearning/meetings/ for updates, also sent to public-webmachinelearning@w3.org 14:32:27 ... I'm aware of a bunch of folks still in process of joining, so expect to have a small group today, also timezones are a challenge 14:33:07 ... today want to discuss three things: 1) high vs. low level API proposed by Greg; 2) use cases review and gauge whether we have consensus to use these as a starting point for the API definition 3) WebNN proof-of-concept update since TPAC 14:33:36 How do I join? It asked for password, but I don't have one. 14:34:09 https://lists.w3.org/Archives/Member/internal-webmachinelearning/2018Nov/0000.html 14:34:28 TOPIC: High level vs low level discussion 14:34:37 Present+ Asim_Hussain 14:34:52 anssik: the group's current charter has in scope a low-level API for inference. Greg raised an issue whether that's the right abstraction level, Greg please introduce the issue 14:34:59 -> https://github.com/webmachinelearning/webnn/issues/3 High level vs low level issue 14:35:25 Yes it asks for user name and password. What should I enter? 14:35:31 Present+ Gabe_Esteven 14:35:38 Present+ Asim_Hussain 14:39:35 gregwhitworth: To try and disambiguate the difference between low and high level - we can have an operator level API or load model API 14:40:13 RRSAgent, draft minutes v2 14:40:13 I have made the request to generate https://www.w3.org/2018/12/13-webmachinelearning-minutes.html anssik 14:42:56 Chris Heilmann: I can see that most of the use cases will be sending data to a pre-trained model for web developers, not doing own training on device. The main thing I am worrying about is performance as in a client-side (browser) JavaScript context, we never can control the machines it runs on. 14:43:08 Gabe: unclear whether the higher level API would mean to communicate with a pre-trained model? 14:43:51 Chris Heilmann: I'd love to be able to have a use case where I can run a model on device and request information from it by sending data to it. The benefits would be avoiding latency to a cloud service, and keeping my data on my device. 14:46:55 gregwhitworth: our view is there could be both a high level API for loading a pre-trained model, and a low level API that allows building a computational graph by common building blocks, operations 14:48:59 TOPIC: Use cases review 14:49:37 anssik: Tomoyuki and Ningxin have put together a proposal for use cases, split into "High-Level Use Cases" aka application-level use cases and "Low-Level Use Cases" aka API-level use cases proposing specific API features. 14:49:42 -> https://github.com/webmachinelearning/webnn/pull/2 Use cases PR 14:49:48 -> https://pr-preview.s3.amazonaws.com/tomoyukilabs/webnn/pull/2.html#usecases Use cases HTML Preview 14:50:25 anssik: my expectation is everyone on the call has seen these use cases, so we are not going to read them aload here. Any specific issues that would benefit from discussion on this call 14:50:47 ... I'd like to start a week-long Call for Consensus for adopting these use cases as a starting point for the API definition. 14:50:47 ... Concrete feedback should be provided as comments in the PR. 14:51:23 rgesteve has joined #webmachinelearning 14:52:03 gregwhitworth: the high level use cases align nicely with our internal discussion 14:53:25 ... few low level use cases seem a bit of a stretch, but not going to push back on those, custom layer is a good one 14:53:40 anssik: silence means consent 14:54:50 TOPIC: WebNN POC update 14:55:30 anssik: those of you at TPAC saw a proof-of-concept WebNN implementation for doing real-time object detection, real-time human pose estimation, and image classification 14:55:38 -> https://docs.google.com/presentation/d/1ZmBg264_DIpMchq7tLhj6fMi19dcUv8I_kB8viwfxYM/edit?usp=sharing WebNN POC update slides 14:57:01 ningxinhu: progress on WebNN since TPAC, this is a low level API we're investigating, based on Chromium. At TPAC we updated the group we have 9 ops supported, investigating ops support across platforms 14:57:12 ... new op is fully connected, important for some models 14:58:29 ... in native API mapping, enhanced MPS implementation with MPSNNGraph 14:59:57 .. POC focused on performance, so collected perf date for ONNX.js on top of WebNN POC 15:00:08 s/.. POC/... POC 15:00:21 https://github.com/intel/webml-polyfill/blob/master/docs/supported_ops.md 15:00:40 [Ningxin explaining ResNet-50 performance data] 15:01:05 ningxinhu: new use case investigation: video semantic segmentation 15:01:39 ... outcome: good segmentation accuracy, performance not yet real-time 15:04:47 TOPIC: Any Other Business 15:05:17 s/TOPIC: Any Other Business// 15:05:36 Rafael: which Windows APIs are used in the WebNN POC 15:07:03 Present+ Yuan_?_(Microsoft) 15:07:47 RRSAgent, draft minutes v2 15:07:47 I have made the request to generate https://www.w3.org/2018/12/13-webmachinelearning-minutes.html anssik 15:09:46 ningxinhu: the current WebNN POC is using clDNN (https://github.com/intel/clDNN) for both Windows and Linux 15:09:55 TOPIC: Any Other Business 15:10:10 anssik: An unofficial draft of a possible call for a participation to an unconfirmed W3C Workshop on Web & Machine Learning 15:10:15 -> https://www.w3.org/2018/11/machine-learning-workshop/ [DRAFT] W3C Workshop on Web & Machine Learning 15:10:31 anssik: Next telcon scheduled for 7 Jan 2019, right after the holiday so might cancel. Suggested agenda topics please post to the list. 15:11:15 s/7 Jan/10 Jan/ 15:11:35 anssik: thank you for calling in, let's get the use cases finalized so that we can start 2019 with API definition 15:11:42 TOPIC: Adjourn 15:11:48 RRSAgent, draft minutes v2 15:11:48 I have made the request to generate https://www.w3.org/2018/12/13-webmachinelearning-minutes.html anssik 15:12:27 belem has left #webmachinelearning