File size: 6,855 Bytes
c2ba1ac
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3547578
 
c2ba1ac
3547578
 
c2ba1ac
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3547578
 
c2ba1ac
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
import os
import argparse
import random
import onnxruntime
import numpy as np

import torch
from torch.nn import functional as F
from torch.utils import data

import cv2
from PIL import Image
from tqdm import tqdm

from utils import input_transform, pad_image, resize_image, preprocess, get_confusion_matrix 

parser = argparse.ArgumentParser(description='HRNet')
parser.add_argument('-m', '--onnx-model', default='',
                    type=str, help='Path to onnx model.')
parser.add_argument('-r', '--root', default='',
                    type=str, help='Path to dataset root.')
parser.add_argument('-l', '--list_path', default='',
                    type=str, help='Path to dataset list.')
parser.add_argument("--ipu", action="store_true", help="Use IPU for inference.")
parser.add_argument("--provider_config", type=str, 
                    default="vaip_config.json", help="Path of the config file for seting provider_options.")
args = parser.parse_args()

INPUT_SIZE = [512, 1024]
NUM_CLASSES = 19
IGNORE_LABEL = 255


class Cityscapes(data.Dataset):
    def __init__(self, 
                 root, 
                 list_path, 
                 num_classes=19,
                 downsample_rate=8,
                 ignore_label=-1):

        self.root = root
        self.list_path = list_path
        self.num_classes = num_classes
        self.downsample_rate = downsample_rate
        
        self.img_list = [line.strip().split() for line in open(root+list_path)]

        self.files = self.read_files()

        self.label_mapping = {-1: ignore_label, 0: ignore_label, 
                              1: ignore_label, 2: ignore_label, 
                              3: ignore_label, 4: ignore_label, 
                              5: ignore_label, 6: ignore_label, 
                              7: 0, 8: 1, 9: ignore_label, 
                              10: ignore_label, 11: 2, 12: 3, 
                              13: 4, 14: ignore_label, 15: ignore_label, 
                              16: ignore_label, 17: 5, 18: ignore_label, 
                              19: 6, 20: 7, 21: 8, 22: 9, 23: 10, 24: 11,
                              25: 12, 26: 13, 27: 14, 28: 15, 
                              29: ignore_label, 30: ignore_label, 
                              31: 16, 32: 17, 33: 18}
    
    def read_files(self):
        files = []
        for item in self.img_list:
            image_path, label_path = item
            name = os.path.splitext(os.path.basename(label_path))[0]
            files.append({
                "img": image_path,
                "label": label_path,
                "name": name,
            })
        return files

    def __len__(self):
        return len(self.files)
        
    def convert_label(self, label, inverse=False):
        temp = label.copy()
        if inverse:
            for v, k in self.label_mapping.items():
                label[temp == k] = v
        else:
            for k, v in self.label_mapping.items():
                label[temp == k] = v
        return label

    def __getitem__(self, index):
        item = self.files[index]
        image = cv2.imread(os.path.join(self.root, item["img"]),
                           cv2.IMREAD_COLOR)
        label = cv2.imread(os.path.join(self.root, item["label"]),
                           cv2.IMREAD_GRAYSCALE)
        label = self.convert_label(label)
        image, label = self.gen_sample(image, label)

        return image.copy(), label.copy()

    def gen_sample(self, image, label):
        label = self.label_transform(label)
        # image = image.transpose((2, 0, 1))

        if self.downsample_rate != 1:
            label = cv2.resize(
                label,
                None,
                fx=self.downsample_rate,
                fy=self.downsample_rate,
                interpolation=cv2.INTER_NEAREST
            )

        return image, label

    def label_transform(self, label):
        return np.array(label).astype('int32')


def run_onnx_inference(ort_session, img):
    """Infer an image with onnx seession

    Args:
        ort_session: Onnx session
        img (ndarray): Image to be infered.

    Returns:
        ndarray: Model inference result.
    """
    pre_img, pad_h, pad_w = preprocess(img)
    # transform chw into hwc format 

    img = np.expand_dims(pre_img, 0)
    img = np.transpose(img, (0,2,3,1))

    ort_inputs = {ort_session.get_inputs()[0].name: img}
    o1 = ort_session.run(None, ort_inputs)[0]
    h, w = o1.shape[-2:]
    h_cut = int(h / INPUT_SIZE[0] * pad_h)
    w_cut = int(w / INPUT_SIZE[1] * pad_w)
    o1 = o1[..., :h - h_cut, :w - w_cut]
    return o1


def testval(ort_session, root, list_path):

    test_dataset = Cityscapes(
                        root=root,
                        list_path=list_path,
                        num_classes=NUM_CLASSES,
                        ignore_label=IGNORE_LABEL,
                        downsample_rate=1)

    testloader = torch.utils.data.DataLoader(
        test_dataset,
        batch_size=1,
        shuffle=False,
        num_workers=4,
        pin_memory=True)

    confusion_matrix = np.zeros(
        (NUM_CLASSES, NUM_CLASSES))
    for index, batch in enumerate(tqdm(testloader)):
        image, label = batch
        image = image.numpy()[0]
        out = run_onnx_inference(ort_session, image) 
        size = label.size()
        # for hwc output
        out = out.transpose(0, 3, 1, 2)
        if out.shape[2] != size[1] or out.shape[3] != size[2]:
            out = torch.from_numpy(out).cpu()
            pred = F.interpolate(
                out, size=size[1:],
                mode='bilinear'
            )

        confusion_matrix += get_confusion_matrix(
            label,
            pred,
            size,
            NUM_CLASSES,
            IGNORE_LABEL)

    pos = confusion_matrix.sum(1)
    res = confusion_matrix.sum(0)
    tp = np.diag(confusion_matrix)
    pixel_acc = tp.sum()/pos.sum()
    mean_acc = (tp/np.maximum(1.0, pos)).mean()
    IoU_array = (tp / np.maximum(1.0, pos + res - tp))
    mean_IoU = IoU_array.mean()

    return mean_IoU, IoU_array, pixel_acc, mean_acc


if __name__ == "__main__":

    onnx_path = args.onnx_model 
    root = args.root
    list_path = args.list_path
    if args.ipu:
        providers = ["VitisAIExecutionProvider"]
        provider_options = [{"config_file": args.provider_config}]
    else:
        providers = ['CUDAExecutionProvider', 'CPUExecutionProvider']
        provider_options = None
    
    ort_session = onnxruntime.InferenceSession(onnx_path, providers=providers, provider_options=provider_options)
   
    mean_IoU, IoU_array, pixel_acc, mean_acc = testval(ort_session, root, list_path) 

    msg = 'MeanIU: {: 4.4f}, Pixel_Acc: {: 4.4f}, Mean_Acc: {: 4.4f}'.format(mean_IoU, \
        pixel_acc, mean_acc)
    print(msg)