Running very slowly

[CurryKd7]

image_path = ("/home/dongjun/zdj/code/CAST-simple/data/cityscapes/leftImg8bit/train/aachen"
"/aachen_000000_000019_leftImg8bit.png")

加载图像

raw_image = np.array(Image.open(image_path).convert("RGB"))

st = time.time()

构建模型

efficientvit_sam = create_sam_model(
name="xl1", weight_url="/home/dongjun/zdj/model/Efficientvit_sam/xl1.pt",
)
efficientvit_sam = efficientvit_sam.cuda().eval()
efficientvit_mask_generator = EfficientViTSamAutomaticMaskGenerator(efficientvit_sam)
masks, feat = efficientvit_mask_generator.generate(raw_image)
class EfficientViTSamAutomaticMaskGenerator(SamAutomaticMaskGenerator):
def init(
self,
model: EfficientViTSam,
points_per_side: int or None = 64,
points_per_batch: int = 128,
pred_iou_thresh: float = 0.86,
stability_score_thresh: float = 0.92,
stability_score_offset: float = 0.7,
box_nms_thresh: float = 0.7,
crop_n_layers: int = 0,
crop_nms_thresh: float = 0.7,
crop_overlap_ratio: float = 512 / 1500,
crop_n_points_downscale_factor: int = 1,
point_grids: list[np.ndarray] or None = None,
min_mask_region_area: int = 25,
output_mode: str = "binary_mask",

The above is my segmentation of a single image and the selection of some parameters. The running time is about 10 seconds or more, and it is not as efficient as SAM. May I ask why?