local_segment/sam3_demo.py

import os
# if using Apple MPS, fall back to CPU for unsupported ops
# os.environ["PYTORCH_ENABLE_MPS_FALLBACK"] = "1"
import numpy as np
import torch
import matplotlib.pyplot as plt
from PIL import Image
import sam3
import os,requests
from sam3 import build_sam3_image_model
from sam3.model.sam3_image_processor import Sam3Processor
from datetime import datetime
# 指定使用第0块GPU
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
sam3_root = os.path.join(os.path.dirname(sam3.__file__), "..")
# select the device for computation
if torch.cuda.is_available():
    device = torch.device("cuda")
elif torch.backends.mps.is_available():
    device = torch.device("mps")
else:
    device = torch.device("cpu")
print(f"using device: {device}")
torch.autocast("cuda", dtype=torch.bfloat16).__enter__()
# turn on tfloat32 for Ampere GPUs (https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices)
if torch.cuda.get_device_properties(0).major >= 8:
    torch.backends.cuda.matmul.allow_tf32 = True
    torch.backends.cudnn.allow_tf32 = True
np.random.seed(3)

def show_mask(mask):
    color = np.array([255, 255, 255, 1])
        
    h, w = mask.shape[-2:]
    mask = mask.astype(np.uint8)
    mask_image =  mask.reshape(h, w, 1) * color.reshape(1, 1, -1)
    # 将值域限制在 [0, 1]，然后转换为 [0, 255] 的 uint8 类型
    mask_image = np.clip(mask_image, 0, 1)  # 确保值在 [0, 1] 范围内
    mask_image_uint8 = (mask_image * 255).astype(np.uint8)
    return mask_image_uint8
def show_masks(image, masks, scores):
    for i, (mask, score) in enumerate(zip(masks, scores)):
        np_arr = show_mask(mask)
        arr_img = Image.fromarray(np_arr)
        arr_img.save(f"sam3_ax_{i}.png")
        
image_url = "https://ossimg.valimart.net/uploads/vali_ai/20260129/176968033473923.png"
raw_image = Image.open(requests.get(image_url, stream=True).raw).convert("RGB")
bpe_path = f"{sam3_root}/sam3/assets/bpe_simple_vocab_16e6.txt.gz"
checkpoint_path = "/root/.cache/modelscope/hub/models/facebook/sam3/sam3.pt"
model = build_sam3_image_model(bpe_path=bpe_path, enable_inst_interactivity=True,load_from_HF=False,checkpoint_path=checkpoint_path)
start_time = datetime.now()
processor = Sam3Processor(model)
inference_state = processor.set_image(raw_image)

input_point = [[296, 543], [150, 543], [296, 683], [150, 683]]
input_label = [1, 0, 1, 1]

masks, scores, logits = model.predict_inst(
    inference_state,
    point_coords=input_point,
    point_labels=input_label,
    multimask_output=True,
)
mask_input = logits[np.argmax(scores), :, :]  # Choose the model's best mask
masks, scores, logits = model.predict_inst(
    inference_state,
    point_coords=input_point,
    point_labels=input_label,
    mask_input=mask_input[None, :, :],
    multimask_output=False,
)
show_masks(raw_image, masks, scores)
end = datetime.now()
time_diff = end - start_time
print(f"执行时间: {time_diff.total_seconds()} 秒")