| import numpy as np |
| import pyarrow as pa |
|
|
| from dora import DoraStatus |
| from ultralytics import YOLO |
|
|
| pa.array([]) |
|
|
| CAMERA_WIDTH = 960 |
| CAMERA_HEIGHT = 540 |
|
|
| model = YOLO("yolov8n.pt") |
|
|
|
|
| class Operator: |
| """ |
| Infering object from images |
| """ |
|
|
| def on_event( |
| self, |
| dora_event, |
| send_output, |
| ) -> DoraStatus: |
| if dora_event["type"] == "INPUT": |
| frame = ( |
| dora_event["value"].to_numpy().reshape((CAMERA_HEIGHT, CAMERA_WIDTH, 3)) |
| ) |
| frame = frame[:, :, ::-1] |
| results = model(frame, verbose=False) |
| boxes = np.array(results[0].boxes.xyxy.cpu()) |
| conf = np.array(results[0].boxes.conf.cpu()) |
| label = np.array(results[0].boxes.cls.cpu()) |
| |
| arrays = np.concatenate((boxes, conf[:, None], label[:, None]), axis=1) |
|
|
| send_output("bbox", pa.array(arrays.ravel()), dora_event["metadata"]) |
|
|
| return DoraStatus.CONTINUE |
|
|
