Skip to content

Latest commit

 

History

History
287 lines (216 loc) · 9.48 KB

EdgeApplicationSDK_FuncSpec_Deserialize.adoc

File metadata and controls

287 lines (216 loc) · 9.48 KB
Raw

Edge Application SDK
Deserialize
Functional Specifications

TOC

1. Change history

Date What/Why

2023/05/26

Initial draft.

2. Terms/Abbreviations

Terms/Abbreviations Meaning

FlatBuffers

Serialization library

3. Reference materials

4. Expected use case

  • Deserialize the inference results and check the content

5. Functional overview/Algorithm

Functional overview

  • You can decode Base64 encoded inference results

  • You can deserialize what’s serialized in "FlatBuffers"

Deserialize function

Files that support deserialization

  • Binary file of inference results

  • Inference results metadata (or similarly formatted json file) that can be gotten by running inference in "Console for AITRIOS"

    • The following is an example of inference results metadata that can be gotten by running inference in "Console for AITRIOS":

{
  "DeviceID": "xxxxx",
  "ModelID": "xxxxx",
  "Image": true,
  "Inferences": [
    {
      "T": "xxxxx",
      "O": "AQAAAA=="
    }
  ],
  "id": "xxxxx",
  "_rid": "xxxxx",
  "_self": "xxxxx",
  "_etag": "xxxxx",
  "_attachments": "xxxxx",
  "_ts": 0
}

The inference results are stored in "O" of "Inferences". The "AQAAAA==" in "O" is Base64 encoded.

Flow

flowchart TD;
    %% definition
    classDef object fill:#FFE699, stroke:#FFD700
    classDef external_service fill:#BFBFBF, stroke:#6b8e23, stroke-dasharray: 10 2
    style legend fill:#FFFFFF,stroke:#000000

    %% impl
    subgraph legend["Legend"]
        process(Processing/User behavior)
        object[Input/output data]:::object
    end
Loading 
flowchart TD
    %% definition
    classDef object fill:#FFE699, stroke:#FFD700
    style sdk fill:#FFFFFF, stroke:#6b8e23, stroke-dasharray: 10 2

    start((Start))
    id1(Prepare a file to deserialize)
    subgraph sdk[Required files]
    data2[Schema file .fbs]:::object
    data3[Binary or json file]:::object
    end
    id2(Create and edit the configuration file for deserialization)
    id3(Run the notebook to deserialize)
    data1[Deserialized file .json]:::object
    finish(((Finish)))

    %% impl
    start --> id1
    id1 --- sdk
    sdk --- id2
    id2 --- id3
    id3 --- data1
    data1 --> finish
Loading

  • Flow details

    1. Prepare a file to deserialize

      • Prepare a schema file for the serialized data, a file to deserialize, and store it in any directory

    2. Create and edit the configuration file for deserialization

      • Create and edit the configuration file configuration.json to configure notebook runtime settings

    3. Run the notebook to deserialize

Sequence

%%{init:{'themeCSS':'text.actor {font-size:18px !important;} .messageText {font-size:18px !important;} .labelText {font-size:18px !important;} .loopText {font-size:18px !important;} .noteText {font-size:18px !important;}'}}%%
sequenceDiagram
  participant User
  participant Dev Container
  participant docker container

  User->>Dev Container: Prepare a file to deserialize
  User->>Dev Container: Create and edit <br> the configuration file <br> for deserialization
  User->>Dev Container: Run the notebook <br> to deserialize
  opt If the file to deserialize is a json file <br> (specified by input_type in the configuration file)
    Dev Container->>Dev Container: Gets Base64 encoded inference results <br> from json file
    Dev Container->>Dev Container: Decodes the Base64 encoded inference results
    Dev Container->>Dev Container: Creates a Base64 decoded <br> binary file
    Dev Container-->>User: Results
  end
  Dev Container->>docker container: Runs a shell script <br> for deserialization
  Note left of docker container:Schema file
  Note left of docker container:Binary file
  docker container->>docker container: Deserializes <br> the binary file
  docker container->>Dev Container: Creates a deserialized <br> json file
  Note right of Dev Container:json file
  Dev Container-->>User: Results
  opt If the file to deserialize <br> is a json file <br> (specified by input_type <br> in the configuration file)
    Dev Container->>Dev Container: Removes the binary file
    Dev Container-->>User: Results
  end
Loading

6. User interface specifications

How to start each function

  1. Jump to the README.md in the deserialize directory from the hyperlink in the README.md in the directory for each feature that uses deserialization

Prepare a file to deserialize

  1. Prepare a schema file for the serialized data, a file to deserialize, and store it in any directory under the deserialize directory

Note
 If you want to use the results of running inference in "Console for AITRIOS", see "Console User Manual" for details.

Create and edit the configuration file for deserialization

Note
 All parameters are required, unless otherwise indicated.
Note
 All values are case sensitive, unless otherwise indicated.
Note
 Do not use symbolic links to files and directories.
Note
 Only paths under the deserialize directory can be specified for directory paths and file paths.

  1. Create and edit the configuration file, configuration.json, in the deserialize directory.

Configuration Meaning Range Remarks

schema_file

Schema file path

Absolute path or relative to the notebook (*.ipynb)

serialized_file

Path of the file to deserialize

Absolute path or relative to the notebook (*.ipynb)

If you omit the setting for output_dir, or specify an empty string, you can not set a file in the same directory as the notebook (*.ipynb).

input_type

Type of file to deserialize

"binary" or "json"

"binary": Binary file of inference results
"json": The metadata json file where the inference results are stored

output_dir

Path to which the file will be output

Destination of the deserialization results file

Absolute path or relative to the notebook (*.ipynb)

Optional
If omitted or empty, the path to which the file will be output is the same directory as the notebook (*.ipynb).

The format of the output filename is as follows:
"File name of serialized_file without extension".json

If the path to which the file will be output contains a file that duplicates the output filename, an error message is displayed and running is interrupted.

Run the notebook to deserialize

  1. Open the notebook, deserialize.ipynb, in the deserialize directory, and run the python scripts in it

    • The scripts do the following:

      • Checks that configuration.json exists in the deserialize directory

        • If an error occurs, the error description is displayed and running is interrupted.

      • Checks the contents of configuration.json

        • If an error occurs, the error description is displayed and running is interrupted.

      • Checks that configuration.json includes values for schema_file and serialized_file

        • If an error occurs, the error description is displayed and running is interrupted.

      • Checks that configuration.json includes value for output_dir

        • If the directory specified by output_dir does not already exist, it is created at the same time.

        • If an error occurs, the error description is displayed and running is interrupted.

      • Gets inference results in json file when input_type in configuration.json is "json"

        • If getting inference results in json file is successful, decodes Base64 encoded inference results

          • If decoding Base64 is successful, outputs Base64 decoded file to the path of output_dir

          • Output filename is "filename of serialized_file without extension".bin and removes it after deserialization is complete

        • If an error occurs, the error description is displayed and running is interrupted.

      • Runs a shell script for deserialization

        • If successful, outputs deserialization results file to the path of output_dir

        • Output filename is "filename of serialized_file without extension".json

        • If an error occurs, the error description is displayed and running is interrupted.

      • Deletes output Base64 decoded file when input_type is "json"

        • If an error occurs, the error description is displayed and running is interrupted.

      • Displays link to deserialization results file

        • Click the link to view the contents of the deserialization results file

7. Target performances/Impact on performances

  • Usability

    • When the SDK environment is built, users can deserialize without any additional installation steps

    • UI response time of 1.2 seconds or less

    • If processing takes more than 5 seconds, indicates that processing is in progress with successive updates

    • Provides users with documentation of usage tools and version information

8. Assumption/Restriction

  • None

9. Remarks

  • None

10. Unconfirmed items

  • None