-
Notifications
You must be signed in to change notification settings - Fork 2
/
a2d_test.py
157 lines (125 loc) · 5.69 KB
/
a2d_test.py
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
import datetime
import os
import math
import sys
import random
import numpy as np
import torch
from hyper_para import HyperParameters
from torch import nn
from model.model import PropagationModel
from dataset.A2D_dataset import A2DTestDataset
from torch.nn import functional as F
from model.network import PropagationNetwork
import matplotlib.pyplot as plt
import pandas as pd
import imageio
from PIL import Image
import cv2
import zipfile
from einops import rearrange
import time
def aggregate_wbg(prob, keep_bg=False):
new_prob = torch.cat([
torch.prod(1-prob, dim=1, keepdim=True),
prob
], 1).clamp(1e-7, 1-1e-7)
logits = torch.log((new_prob /(1-new_prob)))
if keep_bg:
return F.softmax(logits, dim=1)
else:
return F.softmax(logits, dim=1)[1:]
def resize_and_crop(im, input_h, input_w):
# Resize and crop im to input_h x input_w size
im_h, im_w = im.shape[-2:]
scale = max(input_h / im_h, input_w / im_w)
resized_h = int(np.round(im_h * scale))
resized_w = int(np.round(im_w * scale))
crop_h = int(np.floor(resized_h - input_h) / 2)
crop_w = int(np.floor(resized_w - input_w) / 2)
resized_im = F.interpolate(im, [resized_h, resized_w], mode="bilinear")
new_im = torch.zeros((im.shape[0], 1, input_h, input_w))
new_im[...] = resized_im[:, :, crop_h:crop_h+input_h, crop_w:crop_w+input_w]
return new_im
SMOOTH = 1e-6
def calculate_IoU(pred, gt):
IArea = (pred & (gt == 1.0)).astype(float).sum()
OArea = (pred | (gt == 1.0)).astype(float).sum()
IoU = (IArea + SMOOTH) / (OArea + SMOOTH)
return IoU, IArea, OArea
def a2d_evaluate(args, model, data_loader, device, iterations, output_dir):
model.eval()
save_root = os.path.join(output_dir, "{}/".format(iterations))
MeanIoU, IArea, OArea, Overlap = [], [], [], []
with torch.no_grad():
start_time = time.time()
for data in data_loader:
image = data["images"].to(device)
flow = data["flows"].to(device)
gt = data["masks"]
# target_index = data["target_index"]
original_w, original_h = gt.shape[2], gt.shape[1] #(w, h)
sequence_name = data["sequence_name"][0]
sentences_raw = data["sentence_raw"][0]
sentences = data["tensor_embeddings"].to(device)
attentions = data["attention_masks"].to(device)
logits = model.inference(image, flow, sentences, attentions)
pred = torch.sigmoid(logits)
target_pred = pred
target_pred = resize_and_crop(target_pred, original_h, original_w)
target_gt = gt.unsqueeze(0)
target_pred = target_pred.cpu().data.numpy()
target_pred = (target_pred > np.max(target_pred) * 0.5).astype(np.uint8)
target_gt = target_gt.data.numpy()
iou, iarea, oarea = calculate_IoU(target_pred, target_gt)
MeanIoU.append(iou)
IArea.append(iarea)
OArea.append(oarea)
Overlap.append(iou)
print("sequence: ", sequence_name, "frame:", data["frame_name"][0], "sentences_raw: ", sentences_raw)
prec5, prec6, prec7, prec8, prec9 = np.zeros((len(Overlap), 1)), np.zeros((len(Overlap), 1)), np.zeros((len(Overlap), 1)), np.zeros((len(Overlap), 1)), np.zeros((len(Overlap), 1))
for i in range(len(Overlap)):
if Overlap[i] >= 0.5:
prec5[i] = 1
if Overlap[i] >= 0.6:
prec6[i] = 1
if Overlap[i] >= 0.7:
prec7[i] = 1
if Overlap[i] >= 0.8:
prec8[i] = 1
if Overlap[i] >= 0.9:
prec9[i] = 1
mAP_thres_list = list(range(50, 95 + 1, 5))
mAP = []
for i in range(len(mAP_thres_list)):
tmp = np.zeros((len(Overlap), 1))
for j in range(len(Overlap)):
if Overlap[j] >= mAP_thres_list[i] / 100.0:
tmp[j] = 1
mAP.append(tmp.sum() / tmp.shape[0])
end_time = time.time()
total_time = end_time - start_time
return np.mean(np.array(MeanIoU)), np.array(IArea).sum() / np.array(OArea).sum(), prec5.sum() / prec5.shape[0], prec6.sum() / prec6.shape[0], prec7.sum() / prec7.shape[0], prec8.sum() / prec8.shape[0], prec9.sum() / prec9.shape[0], np.mean(np.array(mAP)), total_time
if __name__ == '__main__':
os.environ["CUDA_VISIBLE_DEVICES"] = "5"
para = HyperParameters()
para.parse()
num_gpus = torch.cuda.device_count()
a2d_dataset_root = para["a2d_root"]
dataset_test = A2DTestDataset(para, a2d_dataset_root)
model_path = "model_path"
output_dir = "output_path"
test_sampler = torch.utils.data.SequentialSampler(dataset_test)
data_loader_test = torch.utils.data.DataLoader(dataset_test, batch_size=1, sampler=test_sampler, num_workers=8)
model = PropagationNetwork(para).cuda()
checkpoint = torch.load(model_path, map_location='cpu')
model.load_state_dict(checkpoint)
device = torch.device("cuda")
iterations = int(model_path.split("/")[-1].split("_")[-1][:-4])
mean_iou, overall_iou, precision5, precision6, precision7, precision8, precision9, precision_mAP, total_time = a2d_evaluate(args=para, model=model, data_loader=data_loader_test, device=device, iterations=iterations,
output_dir=output_dir)
print(f'Test split results:\n'
f'Precision@0.5 {precision5:.3f}, Precision@0.6 {precision6:.3f}, '
f'Precision@0.7 {precision7:.3f}, Precision@0.8 {precision8:.3f}, Precision@0.9 {precision9:.3f},\n'
f'mAP Precision @0.5:0.05:0.95 {precision_mAP:.3f},\n'
f'Overall IoU {overall_iou:.3f}, Mean IoU {mean_iou:.3f}， Total Time {total_time:.3f}')