TI EdgeAI Studio Agent

TI EdgeAI Studio Agent is required to be started on a TI Analytics device to use EdgeAI Studio with the device. To know more about EdgeAI Studio please visit https://dev.ti.com/edgeaistudio

Steps to Run

Fetching IP Address of Device via UART Terminal

2. Using UART Terminal

1. Connect the UART cable to PC and open terminal

2. The COM ports(/dev/ttyUSB) can vary. To check all available usb serial ports, use:

   ls /dev/tty | grep USB

   

   [NOTE: TDA4VM,AM68A,AM69A will mostly be on ttyUSB2 while AM62A,AM62X,AM62P are generally on ttyUSB0. Please refer to EdgeAI SDK documentation (Getting Started) for more information.]

3. Start minicom session on PC using:

   sudo minicom -D /dev/ttyUSBX

4. Login using "root"

   

5. Use ifconfig to get the ip address, the highlighted one as shown in the figure below:

   

2. Using a display

1. If you have a disply, you can connect display to the board. When the board boots up, it will show the IP address of the device on the default screen.

B) Running device agent on target

1. Navigate to edgeai-studio-agent folder:

   cd /opt/edgeai-studio-agent

2. Execute the script:

   ./run.sh

API Documentation

GET Requests

GET /sensor (gets all the sensor attached to the target)

Details

Find all the sensors attached to the target. This includes all the USB Camera connected to the target. It also provides other information like SDK version and device type.

Parameters

None

Responses

http code	content-type	response
200	application/json	All attached sensors
404	application/json	{'detail': 'Sensor not found'}

Example GET response (200)

[
    {
        "name": "UVC Camera (046d:0825)",
        "id": "null",
        "type": "Webcam",
        "device": [
            {
                "id": "/dev/video-usb-cam0",
                "type": "V4L2",
                "description": "device available for capture",
                "status": "available"
            }
        ],
        "sdk_version": "9.0",
        "device_name": "TDA4VM"
    }
]

GET /sensor-session (gets all the sensor sessions)

Details

Find all sensor sessions including initiated and started ones. The target device only supports one session at a time.

Parameters

None

Responses

http code	content-type	response
200	application/json	Sensor session
404	application/json	{'detail': 'Session not found'}

Example GET response (200)

{
    "session": {
        "id": "88c046f1-4ef1-456f-9978-c3a20456b05c",
        "http_port": 8080,
        "http_url": "/raw_imagestream",
        "http_status": "started",
        "http_pid": 1231,
        "ws_port": 8000,
        "ws_url": "",
        "ws_status": "down",
        "ws_pid": 0,
        "udp_server_port": 8081,
        "udp_client_port": 0,
        "udp_status": "started",
        "udp_pid": 1231,
        "tcp_server_port": 0,
        "tcp_client_port": 0,
        "tcp_status": "Down",
        "tcp_pid": 0,
        "data_pipeline_status": "down",
        "data_pipeline_pid": 0,
        "stream_type": "null"
    },
    "sensor": {
        "name": "test_name",
        "id": "test_id",
        "type": "test_type",
        "device": [
            {
                "id": "/dev/video-usb-cam0",
                "type": "V4L2",
                "description": "V4L2",
                "status": "available"
            }
        ],
        "sdk_version": "9.0",
        "device_name": "TDA4VM"
    },
    "project": null,
    "inference": null
}

Note: "id": "88c046f1-4ef1-456f-9978-c3a20456b05c" is the unique id of the sensor-session

GET /sensor-session/{id} (gets sensor session by id)

Details

Find the unique sensor sessions by id.

Parameters

name	type	data type	description
id	required	string	The specific unique idendifier

Responses

http code	content-type	response
200	application/json	Sensor session with given id
400	application/json	{'detail': 'Invalid ID supplied'}
404	application/json	{'detail': 'Session not found'}

Example GET response (200)

{
    "session": {
        "id": "88c046f1-4ef1-456f-9978-c3a20456b05c",
        "http_port": 8080,
        "http_url": "/raw_imagestream",
        "http_status": "started",
        "http_pid": 1288,
        "ws_port": 8000,
        "ws_url": "",
        "ws_status": "down",
        "ws_pid": 0,
        "udp_server_port": 8081,
        "udp_client_port": 0,
        "udp_status": "started",
        "udp_pid": 1288,
        "tcp_server_port": 0,
        "tcp_client_port": 0,
        "tcp_status": "Down",
        "tcp_pid": 0,
        "data_pipeline_status": "down",
        "data_pipeline_pid": 0,
        "stream_type": "null"
    },
    "sensor": {
        "name": "test_name",
        "id": "test_id",
        "type": "test_type",
        "device": [
            {
                "id": "/dev/video-usb-cam0",
                "type": "V4L2",
                "description": "V4L2",
                "status": "available"
            }
        ],
        "sdk_version": "9.0",
        "device_name": "TDA4VM"
    },
    "project": null,
    "inference": null
}

GET /project (gets registered project)

Details

Find registered project

Parameters

None

Responses

http code	content-type	response
200	application/json	Project
404	application/json	{'detail': 'Project not found'}

Example GET response (200)

[
    {
        "id": "88c046f1-4ef1-456f-9978-c3a20456b05c",
        "name": "test_project",
        "sensor": "null",
        "task_type": "classification",
        "model": "null",
        "target_device": "null",
        "model_file": "null",
        "model_file_checksum": "null"
    }
]

GET /project/{id} (gets registered project by id)

Details

Find the registered project by id.

Parameters

name	type	data type	description
id	required	string	The specific unique idendifier

Responses

http code	content-type	response
200	application/json	Project with given id
404	application/json	{'detail': 'Project not found'}

Example GET response (200)

[
    {
        "id": "88c046f1-4ef1-456f-9978-c3a20456b05c",
        "name": "test_project",
        "sensor": "null",
        "task_type": "classification",
        "model": "null",
        "target_device": "null",
        "model_file": "null",
        "model_file_checksum": "null"
    }
]

---

POST Requests

POST /sensor-session (initiate a sensor session)

Details

Initiate a unique sensor session by generating a unique id and setting up all the required client and server processes

Example POST data

{
    "name": "test_name",
    "id": "test_id",
    "type": "test_type",
    "device": [
        {
            "id": "/dev/video-usb-cam0",
            "type": "V4L2",
            "description": "device available for capture",
            "status": "available"
        }
    ],
    "sdk_version": "9.0",
    "device_name": "TDA4VM"
}

Responses

http code	content-type	response
202	application/json	Sensor session with generated unique id
405	application/json	{'detail': 'Invalid input'}

POST /project (register a project)

Details

Register a project to the target and set up project entry with all required parameters supplied

Example POST data

{
    "id": "88c046f1-4ef1-456f-9978-c3a20456b05c",
    "name": "test_project",
    "task_type": "classification"
    "sensor": "null"
    "model": "null"
    "target_device": "null"
    "model_file": "null"
    "model_file_checksum": "null"
}

Responses

http code	content-type	response
201	text/html;charset=utf-8	Succesfully created

POST /project/{id}/model (upload model to the target)

Details

Upload model tarball to the target as outputFile.tar.gz. Each call will overwrite the previous file and checksum with the new file.

Parameters

name	type	data type	description
id	required	string	The specific unique idendifier

Example POST data

File object of the model tarball renamed as outputFile.tar.gz

Responses

http code	content-type	response
201	text/html;charset=utf-8	Succesfully created, {id}
404	application/json	{'detail': 'Project not found'}
405	application/json	{'detail': 'Invalid input'}

---

PUT Requests

PUT /sensor-session/{id} (start session and associated data pipeline)

Details

Start sensor session by setting up the required data pipeline (inference or raw capture). Whether to start inference pipeline or raw capture pipeline is decided by the key "inference" in the PUT data. Client creates the project_session object by aggregating the initiated session object, project object and inference parameter. Sensor session object and project object will be validated, data pipeline process will be started for direct sensor data or inference output based on the value of inference parameter. If successfully started, the response will have data_pipeline_status as 'started' and data_pipeline_pid with the PID of the pipeline process supplying data to the UDP/TCP server/client. This API can also produce and stream data over websockets for browser front end application to display on the UI, for e.g., for log streaming, memory usage, inference time etc. If websocket interface is started, the response will have ws_status as 'started', ws_url & ws_port duly filled, and ws_pid with the PID of the websocket server process supplying data to the Javascript front end application.

Parameters

name	type	data type	description
id	required	string	The specific unique idendifier

Example PUT data

{
    "session": {
        "id": "88c046f1-4ef1-456f-9978-c3a20456b05c",
        "http_port": 8080,
        "http_url": "/raw_imagestream",
        "http_status": "started",
        "http_pid": 1999,
        "ws_port": 8000,
        "ws_url": "",
        "ws_status": "down",
        "ws_pid": 0,
        "udp_server_port": 8081,
        "udp_client_port": 0,
        "udp_status": "started",
        "udp_pid": 1999,
        "tcp_server_port": 0,
        "tcp_client_port": 0,
        "tcp_status": "Down",
        "tcp_pid": 0,
        "data_pipeline_status": "down",
        "data_pipeline_pid": 0,
        "stream_type": "image"
    },
    "sensor": {
        "name": "test_name",
        "id": "test_id",
        "type": "test_type",
        "device": [
            {
                "id": "/dev/video-usb-cam0",
                "type": "V4L2",
                "description": "V4L2",
                "status": "available"
            }
        ],
        "sdk_version": "9.0",
        "device_name": "TDA4VM"
    },
    "inference": false
}

Responses

http code	content-type	response
202	application/json	Data with updated datapipeline and websocket status
400	application/json	{'detail': 'Invalid ID supplied'}
404	application/json	{'detail': 'Session not found'}
404	application/json	{'detail': 'Project not found'}
409	application/json	{'detail': 'Sensor session alreday running'}

---

DELETE Requests

DELETE /sensor-session/{id} (deletes sensor session by id)

Details

Delete sensor session with given id. All processes and resources associated to the sensor sessions will be cleared and freed.

Responses

http code	content-type	response
202	text/html;charset=utf-8	Operation accepted
400	application/json	{'detail': 'Invalid ID supplied'}
404	application/json	{'detail': 'Session not found'}

DELETE /sensor-session/{id}/dpipe (deletes data pipeline associated with sensor session)

Details

Deletes only the data pipeline associated with sensor session having the given id. The data pipeline are the inference or raw data capture pipelines running on the target. The websockets interface associated with the data pipelines are also terminated. The same data pipeline associated with session can be later restarted using PUT method on /sensor-session/{id}.

Responses

http code	content-type	response
202	text/html;charset=utf-8	Operation accepted
400	application/json	{'detail': 'Invalid ID supplied'}
404	application/json	{'detail': 'Session not found'}

DELETE /project/{id} (deletes project by id)

Details

Deletes the registered project with given id. All the resources including model files, artifacts and configuration files related to this project will be cleared and freed.

Responses

http code	content-type	response
200	text/html;charset=utf-8	Successful operation
404	application/json	{'detail': 'Project not found'}