update725

1.py

-import numpy as np
-'''
-A= np.array([[1,1,1,1,1,1,1,1,1,1,1],
-                [-5,-4,-3,-2,-1,0,1,2,3,4,5],
-            ]).transpose()
-b= np.array([2,7,9,12,13,14,14,13,10,8,4])
-
-AA = np.dot(A.transpose(),A)
-print(AA)
-AB = np.dot(A.transpose(),b)
-print(AB)
-x = np.dot(np.linalg.inv(AA),AB)
-print(x)
-
-print(np.dot(b-np.dot(A,x),(b-np.dot(A,x).transpose())))
-#[25, 16, 9, 4, 1, 0, 1, 4, 9, 16, 25]
-
-
-A = np.array([[1,3,1,-4],
-             [-1,-3,1,0],
-             [2,6,2,-8]])
-
-AA = np.dot(A.transpose(),A)
-print(AA)
-
-print(np.linalg.matrix_rank(A))
-'''
-
-import matplotlib.pyplot as plt
-import numpy as np
-plt.rcParams['font.sans-serif']=['SimHei']
-plt.rcParams['axes.unicode_minus'] = False
-x = np.random.rand(10000)
-t = np.arange(len(x))
-# plt.plot(t, x, 'g.', label=u'均匀分布')  # 散点图
-plt.hist(x, 1000, color='m', alpha=0.6, label=u'均匀分布', normed=True)
-plt.legend(loc='upper right')
-plt.grid(True, ls=':')
-plt.show()
-
-
-
-
-
-

visualization.py

-import torch
-import torch.nn as nn
-import torch.optim as optim
-
-import cv2
-import numpy as np
-import torchvision
-from torchvision import datasets
-
-# coding: utf-8
-from PIL import Image
-from torch.utils.data import Dataset
-from torchvision import transforms
-from utils import load_model
-
-check = False
-load_pretrained = False
-
-#model = ResNet_aac(fb=64, n_label = 11, model_size=152,kernel_size=3,stride=2,dk=1,dv=1, Nh=8,shape=224,relative = False)
-model = load_model('resnet50_pmg',pretrain=False,require_grad=True,num_class=101)
-model.cuda()
-model = nn.DataParallel(model)
-#model = resnet152()
-pointlist=[]
-
-if load_pretrained:
-    model_dict = model.state_dict()
-    #pretrained_dict = torch.load(args.pretrainedmodel)
-    pretrained_dict = torch.load("D:/resnet50.pth")
-    # pretrained_dict = model_zoo.load_url('https://download.pytorch.org/models/resnet50-19c8e357.pth')
-    state_dict = {}
-
-    for k, v in model_dict.items():
-        # if k in dict1.keys():
-        if k in pretrained_dict.keys() and "fc" not in k:
-            #state_dict[k] = pretrained_dict[dict1[k]]
-            state_dict[k] = pretrained_dict[k]
-
-        else:
-            state_dict[k] = v
-            print(k)
-
-if check:
-    #filename = "best_model_"
-    #checkpoint = torch.load('./checkpoint/' + filename + 'ckpt.t7')
-    #checkpoint = torch.load('unprebest.t7')
-    model.module.load_state_dict(torch.load("./food101/model.pth").module.state_dict())
-
-
-model.eval()
-test_img = "G:/images/apple_pie/116697.jpg"
-img = Image.open(test_img).convert('RGB')
-transform_test = transforms.Compose([
-            transforms.Resize(size=(299, 299)),
-            transforms.CenterCrop((224,224)),
-            transforms.ToTensor(),
-            transforms.Normalize((0.5457954,0.44430383,0.34424934), (0.23273608,0.24383051,0.24237761))
-        ])
-img1 = transform_test(img).reshape([1,3,224,224])
-model(img1,True)
-#print(model(img1,True).shape)
-
-
-def on_EVENT_LBUTTONDOWN(event, x, y, flags, param):
-    if event == cv2.EVENT_LBUTTONDOWN:
-        xy = "%d,%d" % (x, y)
-        print (xy)
-        pointlist.append((x,y))
-        cv2.circle(img, (x, y), 1, (255, 0, 0), thickness = -1)
-        cv2.putText(img, xy, (x, y), cv2.FONT_HERSHEY_PLAIN,
-                    1.0, (0,0,0), thickness = 1)
-        cv2.imshow("Image", img)
-        return
-
-
-test_img = "G:/images/apple_pie/116697.jpg"
-#显示原图
-img = cv2.imread(test_img)
-
-img = cv2.resize(img,(224,224))
-cv2.namedWindow("Image")
-cv2.setMouseCallback("Image", on_EVENT_LBUTTONDOWN)
-img_raw=img.copy()
-cv2.imshow("Image", img)
-
-cv2.waitKey(0)
-
-
-
-
-#显示normalize之后的图
-img = Image.open(test_img).convert('RGB')
-transform_test = transforms.Compose([
-            transforms.Resize(size=(224, 224)),
-            transforms.ToTensor(),
-            transforms.Normalize((0.5457954,0.44430383,0.34424934), (0.23273608,0.24383051,0.24237761))
-        ])
-img1 = transform_test(img)
-img = img1.transpose(0,1).transpose(1,2)
-img = img.numpy()
-img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
-cv2.imshow("Image", img)
-cv2.waitKey (0)
-
-
-
-#显示热力图
-AAC_mat = np.load(r"D:\PMG\visual\matrix14.npy")
-depth = AAC_mat.shape[-1]
-height = int(np.sqrt(depth))
-AAC_mat = np.reshape(AAC_mat,[8,depth,depth])
-#print(AAC_mat.shape)
-
-
-
-isfirst1 = True
-imgs = None
-imgs1 = None
-for item in pointlist:
-    isfirst = True
-    x,y = item
-    x/=224/height
-    y/=224/height
-    mat = AAC_mat[:,int(x*height+y),:]
-    #print(mat.shape)
-    #for i in range(0,8):
-    result = mat
-
-    result = result.reshape([8,height,height])
-    for i in range(0,8):
-        result = mat[i:]
-        img = result
-        #img = cv2.resize(result, (224, 224))
-        #img = img*255
-
-        heatmap = img / np.max(img)
-        heatmap = np.uint8(255 * heatmap)
-        w=heatmap
-        w = cv2.applyColorMap(heatmap, cv2.COLORMAP_JET)  # 转化为jet 的colormap
-        #w = heatmap
-        w = cv2.resize(w,(224,224))
-        #x = img_raw*0.5+w*0.5   # 权重自己定
-        x =w
-        x = x.astype(np.uint8)
-        #print(x.shape)
-        if isfirst:
-            imgs = x
-            isfirst = False
-        else:
-            print(imgs.shape)
-            print(x.shape)
-            imgs = np.hstack([imgs, x])
-        #print(imgs.shape)
-
-    if isfirst1:
-        imgs1 = imgs
-        isfirst1 = False
-    else:
-        imgs1 = np.vstack([imgs1, imgs])
-
-print(imgs1.shape)
-cv2.imshow("mutil_pic", imgs1)
-cv2.waitKey(0)