max_det problem

[aliva-ksu]

Even if i set the max_det value to 3000 for yolov10, it is not showing more than 300 detection per image. my images have around 800 to 1000 objects per image. the same configuration worked well on yolov8. Here is the code for prediction below:

predictions_count = [] for image_path in glob.glob(dataset directory path): model = YOLO(my model path) # build a new model from YAML

results = model(source = image_path,
                conf = 0.25,
                max_det = 3000,
               imgsz = 3200, device = 0)

names = model.names
print(names)
obj_id = list(names)[list(names.values()).index('seed')]
count = results[0].boxes.cls.tolist().count(obj_id)
print(count)

[github-yizhang]

try lower conf? After max_det，the conf will remove boxes which confidence is lower than conf.

[Shisah]

having the same problem, it ignores max_det parameter no matter the value, values lower than 300 also has no effect, detections still capped at 300