注意:python>=3.8, pytorch>=1.7,torchvision>=0.8

Feel free to ask any question. 遇到问题欢迎评论区讨论.

官方教程:

https://github.com/facebookresearch/segment-anything

1 环境配置

1.1 安装主要库:

(1)pip:

有可能出现错误,需要配置好Git。

pip install git+https://github.com/facebookresearch/segment-anything.git

(2)本地安装:

有可能出现错误,需要配置好Git。

git clone git@github.com:facebookresearch/segment-anything.gitcd segment-anything; pip install -e .

(3)手动下载+手动本地安装:

zip文件:

链接:https://pan.baidu.com/s/1dQ--kTTJab5eloKm6nMYrg 提取码:1234 

解压后运行:

cd segment-anything-mainpip install -e .

1.2 安装依赖库:

pip install opencv-python pycocotools matplotlib onnxruntime onnx

matplotlib 3.7.1和3.7.0可能报错

如果报错:pip install matplotlib==3.6.2

1.3 下载权重文件:

下载三个权重文件中的一个,我用的第一个。

  • defaultorvit_h:ViT-H SAM model.
  • vit_l:ViT-L SAM model.
  • vit_b:ViT-B SAM model.

如果下载过慢:

链接:https://pan.baidu.com/s/11wZUcjYWNL6kxOH5MFGB-g提取码:1234

2 使用教程

2.1 根据在图片上选择的点扣出物体

原始图像:

导入依赖库和展示相关的函数:

import cv2import matplotlib.pyplot as pltimport numpy as npfrom segment_anything import sam_model_registry, SamPredictordef show_mask(mask, ax, random_color=False):if random_color:color = np.concatenate([np.random.random(3), np.array([0.6])], axis=0)else:color = np.array([30 / 255, 144 / 255, 255 / 255, 0.6])h, w = mask.shape[-2:]mask_image = mask.reshape(h, w, 1) * color.reshape(1, 1, -1)ax.imshow(mask_image)def show_points(coords, labels, ax, marker_size=375):pos_points = coords[labels == 1]neg_points = coords[labels == 0]ax.scatter(pos_points[:, 0], pos_points[:, 1], color='green', marker='*', s=marker_size, edgecolor='white', linewidth=1.25)ax.scatter(neg_points[:, 0], neg_points[:, 1], color='red', marker='*', s=marker_size, edgecolor='white', linewidth=1.25)

确定使用的权重文件位置和是否使用cuda等:

sam_checkpoint = "F:\sam_vit_h_4b8939.pth"device = "cuda"model_type = "default"

模型实例化:

sam = sam_model_registry[model_type](checkpoint=sam_checkpoint)sam.to(device=device)predictor = SamPredictor(sam)

读取图像并选择抠图点:

image = cv2.imread(r"F:\Dataset\Tomato_Appearance\Tomato_Xishi\images\xs_1.jpg")image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)predictor.set_image(image)input_point = np.array([[1600, 1000]])input_label = np.array([1])plt.figure(figsize=(10,10))plt.imshow(image)show_points(input_point, input_label, plt.gca())plt.axis('on')plt.show()

扣取图像(会同时提供多个扣取结果):

masks, scores, logits = predictor.predict(point_coords=input_point,point_labels=input_label,multimask_output=True,)# 遍历读取每个扣出的结果for i, (mask, score) in enumerate(zip(masks, scores)):plt.figure(figsize=(10,10))plt.imshow(image)show_mask(mask, plt.gca())show_points(input_point, input_label, plt.gca())plt.title(f"Mask {i+1}, Score: {score:.3f}", fontsize=18)plt.axis('off')plt.show()

尝试扣取其他位置:

2.2 扣取图像中的所有物体

官方教程:

https://github.com/facebookresearch/segment-anything/blob/main/notebooks/automatic_mask_generator_example.ipynb

依赖库和函数导入:

from segment_anything import sam_model_registry, SamAutomaticMaskGenerator, SamPredictorimport cv2import matplotlib.pyplot as pltimport numpy as npdef show_anns(anns):if len(anns) == 0:returnsorted_anns = sorted(anns, key=(lambda x: x['area']), reverse=True)ax = plt.gca()ax.set_autoscale_on(False)polygons = []color = []for ann in sorted_anns:m = ann['segmentation']img = np.ones((m.shape[0], m.shape[1], 3))color_mask = np.random.random((1, 3)).tolist()[0]for i in range(3):img[:,:,i] = color_mask[i]ax.imshow(np.dstack((img, m*0.35)))

读取图片:

image = cv2.imread(r"F:\Dataset\Tomato_Appearance\Tomato_Xishi\images\xs_1.jpg")image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

实例化模型:

sam_checkpoint = "F:\sam_vit_h_4b8939.pth"model_type = "default"device = "cuda"sam = sam_model_registry[model_type](checkpoint=sam_checkpoint)sam.to(device=device)

分割并展示(速度有点慢):

mask_generator = SamAutomaticMaskGenerator(sam)masks = mask_generator.generate(image)plt.figure(figsize=(20,20))plt.imshow(image)show_anns(masks)plt.axis('off')plt.show()

2.3 根据文字扣取物体

配置另外一个库:

https://github.com/IDEA-Research/Grounded-Segment-Anything

配置教程:

【图像分割】Grounded Segment Anything根据文字自动画框或分割环境配置和基本使用教程_Father_of_Python的博客-CSDN博客