1. 验证集 -- 批量测试和可视化 2. 测试集 -- 批量测试和可视化
2024-06-13 23:20:48
1.验证集val.txt验证测试结果可视化~
一步到位！！！批量验证集测试！！！
import sys
sys.path.append("..")
sys.path.insert(0, '.')
import argparse
import torch
import torch.nn as nn
from PIL import Image
import numpy as np
import cv2
import time
import os #要导入osimport lib.transform_cv2 as T
from lib.models import model_factory
from configs import set_cfg_from_filefrom tqdm import tqdmtorch.set_grad_enabled(False)
np.random.seed(123)sys.path.append('G:\\addwater0906\\')val_path = "G:\\addwater0906\\val\\" # val验证集的路径
data = []
i = 0for line in open("G:\\addwater0906\\val.txt", "r"): # 设置文件对象并读取每一行文件data.append(line)
for img in data:filename = img.split(",")print(filename[0])a = cv2.imread("G:\\addwater0906\\"+filename[0]) # 原图img的路径b = os.path.split(filename[0])[-1] # 去路径下最后一级cv2.imwrite(val_path + b, a)i += 1print("一共验证集val多少图片:", i)
# args
parse = argparse.ArgumentParser()
#parse.add_argument('--config', dest='config', type=str, default='BiSeNet-master/configs/bisenetv2_city.py',)
parse.add_argument('--config', dest='config', type=str, default='./configs/bisenetv2_city.py',)
# parse.add_argument('--weight-path', type=str, default='BiSeNet-master/res/model_final_v2_city.pth',)
parse.add_argument('--weight-path', type=str, default='./waterv3_model_final.pth',) # 模型的路径！！！！
#parse.add_argument('--weight-path', type=str, default='./res/model_final.pth',)
#parse.add_argument('--img-path', dest='img_path', type=str, default='BiSeNet-master/example.png',)
parse.add_argument('--img_path', dest='img_path', type=str, default= val_path,)
args = parse.parse_args()
cfg = set_cfg_from_file(args.config)palette = np.random.randint(0, 256, (256, 3), dtype=np.uint8)# define model
net = model_factory[cfg.model_type](cfg.n_cats, aux_mode='pred')
net.load_state_dict(torch.load(args.weight_path, map_location='cpu'), strict=False)
net.eval()
# net.cuda()# prepare data
to_tensor = T.ToTensor(mean=(0.3257, 0.3690, 0.3223), # city, rgbstd=(0.2112, 0.2148, 0.2115),
)
dir_path = args.img_pathfor file_name in tqdm(os.listdir(dir_path)):path = val_path + file_name #这里表示其中一张图像的路径im = cv2.imread(path)[:, :, ::-1] # 主要是这里读入的路径# 后面的接着写就行# im = to_tensor(dict(im=im, lb=None))['im'].unsqueeze(0).cuda()im = to_tensor(dict(im=im, lb=None))['im'].unsqueeze(0)# inferencet1 = time.time()out = net(im).squeeze().detach().cpu().numpy()# pred = palette[out]cv2.imwrite('G:\\addwater0906\\val_mask\\' + file_name, out) # mask输出全黑路径
# print('图像误分割大于8%个数：', count1)
# print('图像误分割大于5%的个数：', count2)# cv2.imwrite('output.jpg', pred)
# cv2.namedWindow("output", cv2.WINDOW_NORMAL)
# cv2.imshow("output", pred)
# cv2.waitKey(0)
import matplotlib.pyplot as plt
from matplotlib import gridspec
import numpy as np
import cv2def create_pascal_label_colormap():colormap = np.zeros((256, 3), dtype=int)ind = np.arange(256, dtype=int)for shift in reversed(range(8)):for channel in range(3):colormap[:, channel] |= ((ind >> channel) & 1) << shiftind >>= 3return colormapdef label_to_color_image(label):if label.ndim != 2:raise ValueError('Expect 2-D input label')colormap = create_pascal_label_colormap()if np.max(label) >= len(colormap):raise ValueError('label value too large.')return colormap[label]def vis_segmentation(image, seg_map):"""输入图片和分割 mask 的可视化."""plt.figure(figsize=(15, 5))grid_spec = gridspec.GridSpec(1, 4, width_ratios=[6, 6, 6, 1])plt.subplot(grid_spec[0])plt.imshow(image)plt.axis('off')plt.title('input image')plt.subplot(grid_spec[1])seg_image = label_to_color_image(seg_map).astype(np.uint8)plt.imshow(seg_image)plt.axis('off')plt.title('segmentation map')plt.subplot(grid_spec[2])plt.imshow(image)plt.imshow(seg_image, alpha=0.5)plt.axis('off')plt.title('segmentation overlay')unique_labels = np.unique(seg_map)ax = plt.subplot(grid_spec[3])plt.imshow(FULL_COLOR_MAP[unique_labels].astype(np.uint8), interpolation='nearest')ax.yaxis.tick_right()plt.yticks(range(len(unique_labels)), LABEL_NAMES[unique_labels])plt.xticks([], [])ax.tick_params(width=0.0)plt.grid('off')# plt.imsave('G:\\1\\' + image)# plt.show()LABEL_NAMES = np.asarray(['background', 'water']) # 假设只有两类
FULL_LABEL_MAP = np.arange(len(LABEL_NAMES)).reshape(len(LABEL_NAMES), 1)
FULL_COLOR_MAP = label_to_color_image(FULL_LABEL_MAP)img_path = 'G:\\addwater0906\\val\\' # img原图路径
png_path = 'G:\\addwater0906\\val_mask\\' # mask输出全黑路径
i = 0
for img in os.listdir(img_path):print(img)imgfile = img_path + imgpngfile = png_path + imgnew_path = "G:\\addwater0906\\pltsave\\" + img #保存新的验证集val.txt测试结果可视化路径！！！img = cv2.imread(imgfile, 1)img = img[:, :, ::-1]seg_map = cv2.imread(pngfile, 0)vis_segmentation(img, seg_map)plt.savefig(new_path)plt.close()i +=1print('可视化总数量：',i)
print('Done.')
生成可视化的video!!!
import cv2
import glob
import os
img_root = "G:\\water\\val_plt\\" # 是图片序列的位置
fps = 3 # 可以随意调整视频的帧速率#可以用(*'DVIX')或(*'X264'),如果都不行先装ffmepg
fourcc = cv2.VideoWriter_fourcc(*'XVID')
videoWriter = cv2.VideoWriter('water.avi',fourcc,fps,(1500,500),True)#(1280,720)是图像序列的尺寸，True指保存为彩色视频，false为灰度视频
j = 0
for i in sorted(glob.glob(os.path.join(img_root, '*'))):j = j+1frame = cv2.imread(i)print('-> load {} picture'.format(j))videoWriter.write(frame)
videoWriter.release()
2.对测试集合统一处理（尺寸、位深等）
单张不规则尺寸测试！！！！！！
import sys
sys.path.append("..")
sys.path.insert(0, '.')
import argparse
import torch
import numpy as np
from PIL import ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = Trueimport lib.transform_cv2 as T
from lib.models import model_factory
from configs import set_cfg_from_filefrom PIL import Image
from PIL import ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = Trueimport matplotlib.pyplot as plt
from matplotlib import gridspec
import cv2torch.set_grad_enabled(False)
np.random.seed(123)path = '00.png' # 输入图片路径
mask_path = 'G:\\A_water+bisenet\\0922mask\\' # mask输出路径
plt_path = 'G:\\A_water+bisenet\\0922plt\\' # plt可视化路径
imgFile = Image.open(path).convert('RGB')
newImage = imgFile.resize((1024, 2048), Image.BILINEAR) # resize长度、宽度
path = path.split(".")
path = path[0] + '.png'
newImage.save(path)# args
parse = argparse.ArgumentParser()
parse.add_argument('--config', dest='config', type=str, default='./configs/bisenetv2_city.py',)
parse.add_argument('--weight-path', type=str, default='./waterv5_model_final.pth',) # 模型路径
parse.add_argument('--img_path', dest='img_path', type=str, default=path,) # 测试图片输入路径
args = parse.parse_args()
cfg = set_cfg_from_file(args.config)palette = np.random.randint(0, 256, (256, 3), dtype=np.uint8)# define model
net = model_factory[cfg.model_type](cfg.n_cats, aux_mode='pred')
net.load_state_dict(torch.load(args.weight_path, map_location='cpu'), strict=False)
net.eval()
# net.cuda()# prepare data
to_tensor = T.ToTensor(mean=(0.3257, 0.3690, 0.3223), # city, rgbstd=(0.2112, 0.2148, 0.2115),
)
dir_path = args.img_pathim = cv2.imread(path)[:, :, ::-1] # 主要是这里读入的路径im = to_tensor(dict(im=im, lb=None))['im'].unsqueeze(0)out = net(im).squeeze().detach().cpu().numpy()cv2.imwrite(mask_path + path, out) # 这里表示输出mask图像的路径def create_pascal_label_colormap():colormap = np.zeros((256, 3), dtype=int)ind = np.arange(256, dtype=int)for shift in reversed(range(8)):for channel in range(3):colormap[:, channel] |= ((ind >> channel) & 1) << shiftind >>= 3return colormapdef label_to_color_image(label):if label.ndim != 2:raise ValueError('Expect 2-D input label')colormap = create_pascal_label_colormap()if np.max(label) >= len(colormap):raise ValueError('label value too large.')return colormap[label]def vis_segmentation(image, seg_map):"""输入图片和分割 mask 的可视化."""plt.figure(figsize=(15, 5))grid_spec = gridspec.GridSpec(1, 4, width_ratios=[6, 6, 6, 1])plt.subplot(grid_spec[0])plt.imshow(image)plt.axis('off')plt.title('input image')plt.subplot(grid_spec[1])seg_image = label_to_color_image(seg_map).astype(np.uint8)plt.imshow(seg_image)plt.axis('off')plt.title('segmentation map')plt.subplot(grid_spec[2])plt.imshow(image)plt.imshow(seg_image, alpha=0.7)plt.axis('off')plt.title('segmentation overlay')unique_labels = np.unique(seg_map)ax = plt.subplot(grid_spec[3])plt.imshow(FULL_COLOR_MAP[unique_labels].astype(np.uint8), interpolation='nearest')ax.yaxis.tick_right()plt.yticks(range(len(unique_labels)), LABEL_NAMES[unique_labels])plt.xticks([], [])ax.tick_params(width=0.0)plt.grid('off')# plt.show()LABEL_NAMES = np.asarray(['unlabeled','water']) # 假设只有三类
FULL_LABEL_MAP = np.arange(len(LABEL_NAMES)).reshape(len(LABEL_NAMES), 1)
FULL_COLOR_MAP = label_to_color_image(FULL_LABEL_MAP)img_path = path # 原图
png_path = mask_path + path # mask图（全黑）
plt_path = plt_path + path # plt图img = cv2.imread(img_path, 1)
img = img[:, :, ::-1]
seg_map = cv2.imread(png_path, 0)
vis_segmentation(img, seg_map)plt.savefig(plt_path)
plt.close()
# plt.clf()对测试集先统一成（1）.jpg/png不固定的格式但顺序排列的图片！！！！
批量测试集测试一步到位！！！
import sys
sys.path.append("..")
sys.path.insert(0, '.')
import argparse
import torch
import torch.nn as nn
from PIL import Image
import numpy as np
import cv2
import time
import os #要导入os
import os
import glob
import os.path
from PIL import Image
from PIL import ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = Trueimport lib.transform_cv2 as T
from lib.models import model_factory
from configs import set_cfg_from_filefrom tqdm import tqdm
import os.path
from PIL import Image
from PIL import ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = True
import cv2import matplotlib.pyplot as plt
from matplotlib import gridspec
import numpy as np
import cv2
import ostorch.set_grad_enabled(False)
np.random.seed(123)path = "G:\\00000000000000000\\" # 输入测试图片路径files = os.listdir(path) #会按顺序排列[1,2,3,4，.png]格式
i=0for file in files:original = path + os.sep+ files[i]new = path + os.sep + str(i+1) + ".png"os.rename(original,new)i +=1def Resize(file, outdir, width, height):imgFile = Image.open(file)try:newImage = imgFile.resize((width, height), Image.BILINEAR)newImage.save(os.path.join(outdir, os.path.basename(file)))except Exception as e:print(e)i = 0
a = 0
for file in glob.glob("G:\\00000000000000000\\*.png"): # 输入图片所在的目录*.pngimg = Image.open(file).convert('RGB')img.save(file)print(file)i += 1Resize(file, path, 640, 480) # 新图片存放的目录img1 = Image.open(file)print(img1.getbands()) # ('P',) 这种是有彩色的，而L是没有彩色的print(img1.size)if len(img1.getbands()) == 3:a += 1print('图片总数量：', i)
print('总图像是RGB的数量：', a)files1 = os.listdir(path)
i1 = 0
a1 = 0
c1 = 0
for pic in files1:img = Image.open(os.path.join(path, pic))print(pic)print(img.getbands()) # ('P',) 这种是有彩色的，而L是没有彩色的print(img.size)i1 +=1# Img = np.array(img)# a = np.unique(Img)# print(a) #看像素值if len(img.getbands()) == 3:a1 +=1size = img.sizew = size[0] # 宽度h = size[1] # 高度if w == 640:if h == 480:c1 += 1print('测试前处理再此确认输入图像image的总数量: ', i1)
print('测试前处理再此确认输入图像是RGB的数量：', a1)
print('测试前处理再此确认输入图像是640*480的数量：', c1)# args
parse = argparse.ArgumentParser()
parse.add_argument('--config', dest='config', type=str, default='./configs/bisenetv2_city.py',)
parse.add_argument('--weight-path', type=str, default='./v3_model_final.pth',) # 模型路径
parse.add_argument('--img_path', dest='img_path', type=str, default=path,) # 测试图片输入路径
args = parse.parse_args()
cfg = set_cfg_from_file(args.config)palette = np.random.randint(0, 256, (256, 3), dtype=np.uint8)# define model
net = model_factory[cfg.model_type](cfg.n_cats, aux_mode='pred')
net.load_state_dict(torch.load(args.weight_path, map_location='cpu'), strict=False)
net.eval()
# net.cuda()# prepare data
to_tensor = T.ToTensor(mean=(0.3257, 0.3690, 0.3223), # city, rgbstd=(0.2112, 0.2148, 0.2115),
)
dir_path = args.img_path
count1 = 0
count2 = 0
for file_name in tqdm(os.listdir(dir_path)):path1 = path + file_name # 这里表示输入图像的路径im = cv2.imread(path1)[:, :, ::-1] # 主要是这里读入的路径# 后面的接着写就行# im = to_tensor(dict(im=im, lb=None))['im'].unsqueeze(0).cuda()im = to_tensor(dict(im=im, lb=None))['im'].unsqueeze(0)# inferencet1 = time.time()out = net(im).squeeze().detach().cpu().numpy()cv2.imwrite('G:\\11mask\\' + file_name, out) # 这里表示输出mask图像的路径def create_pascal_label_colormap():colormap = np.zeros((256, 3), dtype=int)ind = np.arange(256, dtype=int)for shift in reversed(range(8)):for channel in range(3):colormap[:, channel] |= ((ind >> channel) & 1) << shiftind >>= 3return colormapdef label_to_color_image(label):if label.ndim != 2:raise ValueError('Expect 2-D input label')colormap = create_pascal_label_colormap()if np.max(label) >= len(colormap):raise ValueError('label value too large.')return colormap[label]def vis_segmentation(image, seg_map):"""输入图片和分割 mask 的可视化."""plt.figure(figsize=(15, 5))grid_spec = gridspec.GridSpec(1, 4, width_ratios=[6, 6, 6, 1])plt.subplot(grid_spec[0])plt.imshow(image)plt.axis('off')plt.title('input image')plt.subplot(grid_spec[1])seg_image = label_to_color_image(seg_map).astype(np.uint8)plt.imshow(seg_image)plt.axis('off')plt.title('segmentation map')plt.subplot(grid_spec[2])plt.imshow(image)plt.imshow(seg_image, alpha=0.7)plt.axis('off')plt.title('segmentation overlay')unique_labels = np.unique(seg_map)ax = plt.subplot(grid_spec[3])plt.imshow(FULL_COLOR_MAP[unique_labels].astype(np.uint8), interpolation='nearest')ax.yaxis.tick_right()plt.yticks(range(len(unique_labels)), LABEL_NAMES[unique_labels])plt.xticks([], [])ax.tick_params(width=0.0)plt.grid('off')# plt.show()LABEL_NAMES = np.asarray(['unlabeled','road','sidewalk']) # 假设只有三类
FULL_LABEL_MAP = np.arange(len(LABEL_NAMES)).reshape(len(LABEL_NAMES), 1)
FULL_COLOR_MAP = label_to_color_image(FULL_LABEL_MAP)img_path = path # 原图
png_path = 'G:\\11mask\\' # mask图（全黑）
v = 0
for img in tqdm(os.listdir(img_path)):# print(img)imgfile = img_path + imgpngfile = png_path + imgnew_path = 'G:\\pltsave\\' + img # 可视化测试结果路径img = cv2.imread(imgfile, 1)img = img[:, :, ::-1]seg_map = cv2.imread(pngfile, 0)vis_segmentation(img, seg_map)plt.savefig(new_path)plt.close()v += 1# plt.clf()print('Done！！！！测试集测试完毕并可视化数量：', v)
输出结果如下：
也可以参考分布式处理（不推荐）：
import os
import glob
import os.path
from PIL import Image
from PIL import ImageFile #批量修改尺寸2048*1024（此时也可能32位）
ImageFile.LOAD_TRUNCATED_IMAGES = Truepath = "G:\\000\\image_ground150"
files = os.listdir(path) #会按顺序排列1,2,3,4，.png格式
i=0
for file in files:original = path + os.sep+ files[i]new = path + os.sep + str(i+1) + ".png"os.rename(original,new)i +=1
def Resize(file, outdir, width, height):imgFile = Image.open(file)try:newImage = imgFile.resize((width, height), Image.BILINEAR)newImage.save(os.path.join(outdir, os.path.basename(file)))except Exception as e:print(e)i = 0
a = 0
for file in glob.glob("G:\\000\\image_ground150\\*.png"): # 图片所在的目录img = Image.open(file).convert('RGB')print(file)i += 1Resize(file, path, 640, 480) # 新图片存放的目录img1 = Image.open(file)print(img1.getbands()) # ('P',) 这种是有彩色的，而L是没有彩色的print(img1.size)if len(img1.getbands()) == 3:a +=1
print('图片总数量：', i)
print('总图像是RGB的数量：', a)
检查下是否转化成功：
import os.path
from PIL import Image
from PIL import ImageFile #批量修改尺寸2048*1024（此时也可能32位）
ImageFile.LOAD_TRUNCATED_IMAGES = True
import cv2path = "G:\\000\\image_ground150\\"files = os.listdir(path)
print(files)
i = 0
a = 0
c = 0
for pic in files:img = Image.open(os.path.join(path, pic))print(pic)print(img.getbands()) # ('P',) 这种是有彩色的，而L是没有彩色的print(img.size)i +=1# Img = np.array(img)# a = np.unique(Img)# print(a) #看像素值if len(img.getbands()) == 3:a +=1size = img.sizew = size[0] # 宽度h = size[1] # 高度if w == 640:if h == 480:c += 1print('图像image的总数量: ', i)
print('总图像是RGB的数量：', a)
print('总图像是640*480的数量：', c)
3.demo.py进行测试生成全黑mask
import sys
sys.path.append("..")
sys.path.insert(0, '.')
import argparse
import torch
import torch.nn as nn
from PIL import Image
import numpy as np
import cv2
import time
import os #要导入osimport lib.transform_cv2 as T
from lib.models import model_factory
from configs import set_cfg_from_filefrom tqdm import tqdmtorch.set_grad_enabled(False)
np.random.seed(123)# args
parse = argparse.ArgumentParser()
parse.add_argument('--config', dest='config', type=str, default='./configs/bisenetv2_city.py',)
parse.add_argument('--weight-path', type=str, default='./v3_model_final.pth',)
parse.add_argument('--img_path', dest='img_path', type=str, default='G:\\0907ground_test\\20210906_images\\',)
args = parse.parse_args()
cfg = set_cfg_from_file(args.config)palette = np.random.randint(0, 256, (256, 3), dtype=np.uint8)# define model
net = model_factory[cfg.model_type](cfg.n_cats, aux_mode='pred')
net.load_state_dict(torch.load(args.weight_path, map_location='cpu'), strict=False)
net.eval()
# net.cuda()# prepare data
to_tensor = T.ToTensor(mean=(0.3257, 0.3690, 0.3223), # city, rgbstd=(0.2112, 0.2148, 0.2115),
)
dir_path = args.img_path
count1 = 0
count2 = 0
for file_name in tqdm(os.listdir(dir_path)):path = 'G:\\0907ground_test\\20210906_images\\' + file_name # 这里表示输入图像的路径im = cv2.imread(path)[:, :, ::-1] # 主要是这里读入的路径# 后面的接着写就行# im = to_tensor(dict(im=im, lb=None))['im'].unsqueeze(0).cuda()im = to_tensor(dict(im=im, lb=None))['im'].unsqueeze(0)# inferencet1 = time.time()out = net(im).squeeze().detach().cpu().numpy()# if out.__contains__(1):# count1 += 1# pred = palette[out]# if np.count_nonzero(out > 0) > 24576:# count1 += 1# pred = palette[out]# cv2.imwrite('G:\\000\\ground_8%\\' + file_name, pred)## if np.count_nonzero(out > 0) > 15300:# count2 += 1# pred = palette[out]# cv2.imwrite('G:\\000\\ground_5%\\' + file_name, pred)# pred = palette[out]cv2.imwrite('G:\\0907ground_test\\0907_v3mask\\' + file_name, out) # 这里表示输出图像的路径
# print('图像误分割大于8%个数：', count1)
# print('图像误分割大于5%的个数：', count2)# cv2.imwrite('output.jpg', pred)
# cv2.namedWindow("output", cv2.WINDOW_NORMAL)
# cv2.imshow("output", pred)
# cv2.waitKey(0)
4. 测试预测图像可视化（原图、全黑mask）
import matplotlib.pyplot as plt
from matplotlib import gridspec
import numpy as np
import cv2
import osdef create_pascal_label_colormap():colormap = np.zeros((256, 3), dtype=int)ind = np.arange(256, dtype=int)for shift in reversed(range(8)):for channel in range(3):colormap[:, channel] |= ((ind >> channel) & 1) << shiftind >>= 3return colormapdef label_to_color_image(label):if label.ndim != 2:raise ValueError('Expect 2-D input label')colormap = create_pascal_label_colormap()if np.max(label) >= len(colormap):raise ValueError('label value too large.')return colormap[label]def vis_segmentation(image, seg_map):"""输入图片和分割 mask 的可视化."""plt.figure(figsize=(15, 5))grid_spec = gridspec.GridSpec(1, 4, width_ratios=[6, 6, 6, 1])plt.subplot(grid_spec[0])plt.imshow(image)plt.axis('off')plt.title('input image')plt.subplot(grid_spec[1])seg_image = label_to_color_image(seg_map).astype(np.uint8)plt.imshow(seg_image)plt.axis('off')plt.title('segmentation map')plt.subplot(grid_spec[2])plt.imshow(image)plt.imshow(seg_image, alpha=0.7)plt.axis('off')plt.title('segmentation overlay')unique_labels = np.unique(seg_map)ax = plt.subplot(grid_spec[3])plt.imshow(FULL_COLOR_MAP[unique_labels].astype(np.uint8), interpolation='nearest')ax.yaxis.tick_right()plt.yticks(range(len(unique_labels)), LABEL_NAMES[unique_labels])plt.xticks([], [])ax.tick_params(width=0.0)plt.grid('off')# plt.savefig("G:\\0907ground_test\\plt\\" + image)# plt.show()LABEL_NAMES = np.asarray(['unlabeled','road','sidewalk']) # 假设只有三类
FULL_LABEL_MAP = np.arange(len(LABEL_NAMES)).reshape(len(LABEL_NAMES), 1)
FULL_COLOR_MAP = label_to_color_image(FULL_LABEL_MAP)img_path = 'G:\\0907ground_test\\20210906_images\\' # 原图
png_path = 'G:\\0907ground_test\\0907_v3mask\\' # mask图（全黑）
for img in os.listdir(img_path):print(img)imgfile = img_path + imgpngfile = png_path + img
# imgfile = 'G:\\0907ground_test\\20210906_images\\108.png'
# pngfile = 'G:\\0907ground_test\\0907_v3mask\\108.png'new_path = 'G:\\0907ground_test\\plt\\' + imgimg = cv2.imread(imgfile, 1)img = img[:, :, ::-1]seg_map = cv2.imread(pngfile, 0)vis_segmentation(img, seg_map)plt.savefig(new_path)plt.close()# plt.clf()print('Done.')
1. 验证集 -- 批量测试和可视化 2. 测试集 -- 批量测试和可视化相关推荐
- ubuntu16.04下使用YOLOV3训练自己做的VOC数据集（VOC数据集制作+模型训练步骤+训练参数解析和问题解决+训练日志可视化（loss、IOU）+模型测试（单张、摄像头、批量测试））
前序 1.环境配置 请自行参考其他博客 本机环境 ubuntu16.04 python3 英伟达显卡驱动:nvidia-396 OpenCV3.4.5 CUDNN7.0.5 CUDA9.0 2.ubu ...
- python可视化多个机器学习模型在独立测试集（test data set）上面的AUC值、可视化模型效能
python可视化多个机器学习模型在独立测试集(test data set)上面的AUC值.可视化模型效能 # x_lables为模型名称列表,包括,逻辑回归.朴素贝叶斯.支持向量机.随机森林.xgb ...
- 为什么引入验证集来评估机器学习模型？只用训练集和测试集可以吗？
评估模型的重点是将数据划分为三个集合:训练集.验证集和测试集.在训练数据上训练模型,在验证数据上评估模型.一旦找到了最佳参数,就在测试数据上最后测试一次.你可能会问,为什么不是两个集合:一个训练集和一 ...
- 故障转移集群无法连接到节点_Redis集群以及自动故障转移测试
在Redis中,与Sentinel(哨兵)实现的高可用相比,集群(cluster)更多的是强调数据的分片或者是节点的伸缩性,如果在集群的主节点上加入对应的从节点,集群还可以自动故障转移,因此相比Sen ...
- weka使用训练集分类测试集_Giao 13C NMR计算分类训练集提高结构归属的准确性和可靠性...
Giao 13C NMR计算分类训练集提高结构归属的准确性和可靠性 前言 GIAO 13C NMR计算对于小分子结构归属有很重要的作用,尤其对于需要提高准确性和可信性的复杂有机分子的结构归属.准确度是 ...
- 面经合集（包含python、mysql、linux、测试等内容）
mysql 1. 怎么排序? order by. 2. 聚合函数 sum,count,min,max... 3. 介绍一下mysql 关系型数据库.由于体积小.速度快.成本低.开源,MySQL被广泛的 ...
- 6759字25个APP测试流程以及资料合集【乐搏TestPro】
App测试流程及资料合集 / 1 / 测试理论知识:跟其他web等测试一样: 如果基础不够建议先看书:给软件测试从业者推荐6本书籍 / 2 / 整体测试流程:同web类似 具体差异化见下方内容 / 3 ...
- 渗透测试 ( 5 ) --- 扫描之王 nmap、渗透测试工具实战技巧合集
Nmap 官方文档 ( 中文文档是 Nmap 版本4.50,英文文档是最新的 ): 英文文档:https://nmap.org/book/man.html 中文文档:https://nmap.org/ ...
- linux磁盘fio压力测试,fio命令 – 对磁盘进行压力测试和验证
fio是测试IOPS的非常好的工具,用来对磁盘进行压力测试和验证.磁盘IO是检查磁盘性能的重要指标,可以按照负载情况分成照顺序读写,随机读写两大类.FIO是一个可以产生很多线程或进程并执行用户指定的特 ...
- Echarts+D3可视化开发工具集
Echarts+D3可视化开发工具集 一.前言 大学三年期间,学的很杂,但唯一坚持的就是在做可视化开发,有个负责任的指导老师很重要.因为即将大四,之后不一定会再继续朝这个方面学习,所以把之前参加比 ...
最新文章
- 使用Ranger对Hive数据进行脱敏
- PHP警告:socket_bind():无法绑定地址
- Django的View（视图）
- C和指针之函数之把数字字符串转为整数并且返回这个数字(ascii_to_integer)
- 【洛谷】P1641 [SCOI2010]生成字符串（思维+组合+逆元）
- linux系统核心已存储,Linux系统 存储
- idea server日志乱码_windows下，Kiwi_Syslog日志服务器的搭建
- spring思想分析
- LeetCode84 Largest Rectangle in Histogram
- Linux的环境变量.bash_profile .bashrc profile文件
- 颜色直方图的计算、显示、处理、对比及反向投影(How to Use Histogram? Calculate, Show, Process, Compare and BackProject)...
- 昨天晚上全新打造N无线AP
- 《JAVA 技术》第二次作业
- 你的才艺怎样变现？--Rarible平台
- php 通过API接口连接12306余票查询
- 【转】一起做RGB-D SLAM (1)
- ds12c887程序C语言,时钟芯片DS12C887使用说明详解.doc
- 脑洞文之去火星搞IT!
- 关于BeanUtils.populate()方法使用时的报错解决
- 猿大师播放器有哪些不同的版本，分别适合哪些不同需求的场景？