Laplace_Landmark_Localization/testFace300W80.py at master · FunkyKoki/Laplace_Landmark_Localization · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
import tqdm
import copy
import torch
import numpy as np
import cv2
from models import Generator, loadWeights, GeneratorGhostConv
from datasets import Face300W80, datasetSize
import time
from thop import profile, clever_format

import os
os.environ["CUDA_VISIBLE_DEVICES"] = "0"

def testFace300W80(split, test_epoch, logName, rat=1, tempera=2.0):
    testset = Face300W80(split=split, augment=False)
    dataloader = torch.utils.data.DataLoader(testset, batch_size=1, shuffle=False, pin_memory=True)

#     inputs = torch.randn(1, 3, 80, 80)
#     flops, params = profile(Generator(ratio=1, temperature=1.), inputs=(inputs, ))
#     flops, params = profile(GeneratorGhostConv(ratio=1, temperature=1.), inputs=(inputs, ))
#     flops, params = clever_format([flops, params], "%.3f")
#     print(flops, params)

#     model = GeneratorGhostConv(ratio=rat, temperature=tempera)
    model = Generator(ratio=rat, temperature=tempera)
    model = loadWeights(model, '/media/WDC/savedModels/LaplaceLandmarkLocalization/'+logName+'_model_'+str(test_epoch)+'.pth', 'cuda:0')
    model.eval().cuda('cuda:0')

    errorRates = []
    times = []
    with torch.no_grad():
        for data in tqdm.tqdm(dataloader):
            img, paramsTarget, _, _, _ = data
            img = img.cuda('cuda:0').float()
            paramsTarget = paramsTarget.squeeze().numpy()

            startTime = time.time()
            paramsPredict, _ = model(img)
            times.append(time.time()-startTime)
            paramsPredict = paramsPredict.cpu().squeeze().numpy()

            pts = paramsPredict[:, 0:2]
            tpts = paramsTarget[:, 0:2]

#             normalizeFactor = np.sqrt(((tpts[39, 0]+tpts[36, 0])/2 - (tpts[42, 0]+tpts[45, 0])/2)**2 + ((tpts[39, 1]+tpts[36, 1])/2 - (tpts[42, 1]+tpts[45, 1])/2)**2)
            normalizeFactor = np.sqrt((tpts[36, 0] - tpts[45, 0])**2 + (tpts[36, 1] - tpts[45, 1])**2)

            errorRate = np.sum(np.sqrt(np.sum(pow(pts-tpts, 2), axis=1)))/68/normalizeFactor
            # if errorRate > 0.03:
                # showImgLandmarksInGetItem(imgRaw, pts, name=imgName, oneByOne=False, KPTNUM=19)
                # showImgGtAndPredLandmarks(imgRaw, tpts, pts, name=imgName, oneByOne=False, KPTNUM=19)

            errorRates.append(errorRate)

    errorRate = sum(errorRates) / datasetSize["300W80"][split] * 100
    print(split + ' error rate: ' + str(errorRate))
    print("Avg forward time is: " + str(sum(times)/datasetSize["300W80"][split]) + "ms")
    print("FPS: " + str(1/sum(times)*datasetSize["300W80"][split]))
    return errorRate


def testFace300W80CPU(split, test_epoch, logName):
    testset = Face300W80(split=split, augment=False)
    dataloader = torch.utils.data.DataLoader(testset, batch_size=1, shuffle=False, pin_memory=True)

#     inputs = torch.randn(1, 3, 80, 80)
#     flops, params = profile(Generator(ratio=1, temperature=1.), inputs=(inputs, ))
#     flops, params = profile(GeneratorGhostConv(ratio=1, temperature=1.), inputs=(inputs, ))
#     flops, params = clever_format([flops, params], "%.3f")
#     print(flops, params)

#     model = Generator(ratio=1, temperature=2.)
    model = GeneratorGhostConv(ratio=1, temperature=2.)
    model = loadWeights(model, '/media/WDC/savedModels/LaplaceLandmarkLocalization/'+logName+'_model_'+str(test_epoch)+'.pth', 'cpu')
    model.eval()

    errorRates = []
    times = []
    with torch.no_grad():
        for data in tqdm.tqdm(dataloader):
            img, paramsTarget, _, _, _ = data
            img = img.float()
            paramsTarget = paramsTarget.squeeze().numpy()

            startTime = time.time()
            paramsPredict, _ = model(img)
            times.append(time.time()-startTime)
            paramsPredict = paramsPredict.squeeze().numpy()

            pts = paramsPredict[:, 0:2]
            tpts = paramsTarget[:, 0:2]

            normalizeFactor = np.sqrt((tpts[36, 0] - tpts[45, 0])**2 + (tpts[36, 1] - tpts[45, 1])**2)

            errorRate = np.sum(np.sqrt(np.sum(pow(pts-tpts, 2), axis=1)))/68/normalizeFactor
            # if errorRate > 0.03:
                # showImgLandmarksInGetItem(imgRaw, pts, name=imgName, oneByOne=False, KPTNUM=19)
                # showImgGtAndPredLandmarks(imgRaw, tpts, pts, name=imgName, oneByOne=False, KPTNUM=19)

            errorRates.append(errorRate)

    errorRate = sum(errorRates) / datasetSize["300W80"][split] * 100
    print(split + ' error rate: ' + str(errorRate))
    print("Avg forward time is: " + str(sum(times)/datasetSize["300W80"][split]) + "ms")
    print("FPS: " + str(1/sum(times)*datasetSize["300W80"][split]))
    return errorRate


if __name__ == "__main__":
#     logName = '20200606_face300W80GAN_Train_3'
#     logName = '20200525_face300W80_Train_2'
    logName = '20200523_face300W80_Train_5'
    testFace300W80("full", 'best', logName)