Skip to content

Instantly share code, notes, and snippets.

View samuelgoto's full-sized avatar
💭
hello world!

Sam Goto samuelgoto

💭
hello world!
159 followers · 4 following
View GitHub Profile
@samuelgoto
samuelgoto / rows.md
Last active July 17, 2018 00:33

Discovery

<link rel="alternate" href="/api" type="application/vnd.microforms">

Self-described payloads

{
@samuelgoto
samuelgoto / english.md
Created June 22, 2018 16:57
@samuelgoto
samuelgoto / forms.jsonld.md
Created June 1, 2018 16:08

JSON-LD encapsulation of forms:

{
  @context: "https://example.com",
  @type: MyType,
  action: {
    @context: "https://w3c.org/2018/forms",
    @type: Form,
 ...
@samuelgoto
samuelgoto / deeplearning.jsonld.md
Created June 1, 2018 16:06

CoreML/ONNX/TF JSON-LD based for serving.

{
  @context: "https://w3c.org/2018/deeplearning",
  @type: Model,
  network: {
    @type: NeuralNetwork,
    ...
 }
@samuelgoto
samuelgoto / atom.jsonld.md
Last active August 8, 2018 00:12

ATOM feeds in JSON-LD.

{
  "@context": "http://www.w3.org/2005/Atom",
  "@type": "Feed",
  "title": "Example Feed",
  "subtitle": "A subtitle.",
  "links": [{
@samuelgoto
samuelgoto / arpub.md
Created May 31, 2018 23:17

Alternatives considered

TODO(goto): go over this.

ARML is an XML-based data format to describe an interact with AR scenes.

ARML




  


  




    


          
    

      

        
        

          

              @samuelgoto
          

          

            
                samuelgoto
                / WebDPU.md
            
            

              Last active
              May 29, 2018 20:37
            

          

        

      

        

  

      

    
Deep Learning Processing Units




10-15% cost of interpreation seems like a non-starter to me when they are trying to squeeze in every level of performance.
I'm seeing some convergence in the industry regarding "inference models" file formats / representations




I would challenge that assertion. Your going to get a larger drop on an android phone when thermal throttling kicks in.






Prior Art




  


  




    


          
    

      

        
        

          

              @samuelgoto
          

          

            
                samuelgoto
                / webdl.md
            
            

              Last active
              May 17, 2018 19:07
            

          

        

      

        

  

      

    
    

  • web assembly will always have a 10-12% interpretation cost (anecdotal data from @bnelson, need to sanity check / validate with others), which for computer vision things over a came stream seems like a non starter.
    • 

  • composition, portability
    • 

  • ergonomics (don't have to care about gpu)
    • 

  • cpu vs gpu: memory constrained versus computing contrained
    • 

  • concurrency, multithreading: don't you need to transfer video streams back and fourth between worker threads?
    • 




  


  




    


          
    

      

        
        

          

              @samuelgoto
          

          

            
                samuelgoto
                / script-policies.md
            
            

              Last active
              April 18, 2018 19:39
            

          

        

      

        

  

      

    
Things that could be interesting to control/censor:



Things that "use strict" control:



  


  




    


          
    

      

        
        

          

              @samuelgoto
          

          

            
                samuelgoto
                / forms.md
            
            

              Last active
              April 16, 2018 19:15