I have a project where I am implementing the Yolo object detection algorithm with different tracking algorithms. I am now struggling to implement the StrongSort tracking with my detection program. Can someone explain how I should implement the algorithm with the use of a GitHub repo. How can I feed the detection results into the tracking algorithm? This is my detection algorithm:

from ultralytics import YOLO

class YoloDetector:
  def __init__(self, model_path, confidence):
    self.model = YOLO(model_path)
    self.classList = ["person"]
    self.confidence = confidence

  def detect(self, image):
    results = self.model.predict(image, conf=self.confidence)
    result = results[0]
    detections = self.make_detections(result)
    return detections

  def make_detections(self, result):
    boxes = result.boxes
    detections = []
    for box in boxes:
      x1, y1, x2, y2 = box.xyxy[0]
      x1, y1, x2, y2 = int(x1), int(y1), int(x2), int(y2)
      w, h = x2 - x1, y2 - y1
      class_number = int(box.cls[0])

      if result.names[class_number] not in self.classList:
        continue
      conf = box.conf[0]
      detections.append((([x1, y1, w, h]), class_number, conf))
    return detections

I have a project where I am implementing the Yolo object detection algorithm with different tracking algorithms. I am now struggling to implement the StrongSort tracking with my detection program. Can someone explain how I should implement the algorithm with the use of a GitHub repo. How can I feed the detection results into the tracking algorithm? This is my detection algorithm:

from ultralytics import YOLO

class YoloDetector:
  def __init__(self, model_path, confidence):
    self.model = YOLO(model_path)
    self.classList = ["person"]
    self.confidence = confidence

  def detect(self, image):
    results = self.model.predict(image, conf=self.confidence)
    result = results[0]
    detections = self.make_detections(result)
    return detections

  def make_detections(self, result):
    boxes = result.boxes
    detections = []
    for box in boxes:
      x1, y1, x2, y2 = box.xyxy[0]
      x1, y1, x2, y2 = int(x1), int(y1), int(x2), int(y2)
      w, h = x2 - x1, y2 - y1
      class_number = int(box.cls[0])

      if result.names[class_number] not in self.classList:
        continue
      conf = box.conf[0]
      detections.append((([x1, y1, w, h]), class_number, conf))
    return detections

• python
• object-detection
• tracking
• yolo
• ultralytics

Share Improve this question Follow asked Mar 25 at 15:12 ofhgofofhgof 1

Add a comment  | 

1 Answer 1

Sorted by: Reset to default 0

class YoloDetectorWithTracking:
    def __init__(self, model_path, confidence):
        self.model = YOLO(model_path)
        self.classList = ["person"]
        self.confidence = confidence
        self.tracker = StrongSort()

    def detect_and_track(self, image):
        detections = self.detect(image)
        boxes = np.array([det[0] for det in detections])
        scores = np.array([det[2] for det in detections])
        tracks = self.tracker.update(boxes, scores, image)
        return tracks

    def detect(self, image):
        results = self.model.predict(image, conf=self.confidence)
        result = results[0]
        detections = self.make_detections(result)
        return detections

    def make_detections(self, result):
        boxes = result.boxes
        detections = []
        for box in boxes:
            x1, y1, x2, y2 = box.xyxy[0]
            x1, y1, x2, y2 = int(x1), int(y1), int(x2), int(y2)
            w, h = x2 - x1, y2 - y1
            class_number = int(box.cls[0])

            if result.names[class_number] not in self.classList:
                continue
            conf = box.conf[0]
            detections.append((([x1, y1, w, h]), class_number, conf))
        return detections

I have added the StrongSort tracker to the YOLO detection class. The YOLO detections (bounding boxes and confidence scores) are passed to the tracker, which updates and returns the tracked objects with IDs. These tracked objects are then visualized with bounding boxes and IDs in the video frames.