File size: 6,419 Bytes
57db94b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
214
215
import torch
import torch.nn.functional as F
import cv2
import numpy as np
from torchvision import transforms
import os
import matplotlib.pyplot as plt
from PIL import Image
import time
from skimage.metrics import structural_similarity as ssim
from skimage.color import rgb2lab


from combined import IFNet, warp


device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(f"Using device: {device}")
model = IFNet().to(device)


checkpoint_path = "save_checkpoints/model_epoch_50.pth"
checkpoint = torch.load(checkpoint_path, weights_only=False, map_location=device)
model.load_state_dict(checkpoint['model_state_dict'])
model.eval()

print(f"Loaded model from epoch {checkpoint['epoch']} with PSNR: {checkpoint.get('psnr', 'N/A')} dB")


transform = transforms.Compose([
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])
])

def preprocess_images(img0_path, img1_path, gt_path=None):
    # Read images
    img0 = cv2.imread(img0_path)
    img1 = cv2.imread(img1_path)
    
    if img0 is None or img1 is None:
        raise ValueError(f"Could not read images: {img0_path}, {img1_path}")
    
    
    gt = None
    if gt_path and os.path.exists(gt_path):
        gt = cv2.imread(gt_path)
        if gt is None:
            print(f"Warning: Could not read ground truth image: {gt_path}")
            gt = None
        else:
            gt = cv2.cvtColor(gt, cv2.COLOR_BGR2RGB)
    
    
    img0 = cv2.cvtColor(img0, cv2.COLOR_BGR2RGB)
    img1 = cv2.cvtColor(img1, cv2.COLOR_BGR2RGB)
    
    original_size = (img0.shape[0], img0.shape[1])

    orig_img0 = img0.copy()
    orig_img1 = img1.copy()
    
    img0_resized = cv2.resize(img0, (256, 256))
    img1_resized = cv2.resize(img1, (256, 256))

    img0_tensor = transform(img0_resized)
    img1_tensor = transform(img1_resized)
    

    input_tensor = torch.cat((img0_tensor, img1_tensor), 0).unsqueeze(0).to(device)
    
    return input_tensor, original_size, orig_img0, orig_img1, gt

def tensor_to_image(tensor):
    tensor = tensor.cpu()
    
    tensor = tensor * 0.5 + 0.5
    tensor = tensor.clamp(0, 1)
    
    img = tensor.numpy().transpose(1, 2, 0) * 255
    return img.astype(np.uint8)


def calculate_psnr(img1, img2):
    mse = np.mean((img1.astype(np.float32) - img2.astype(np.float32)) ** 2)
    if mse == 0:
        return float('inf')
    return 10 * np.log10(255.0 ** 2 / mse)

def calculate_ssim(img1, img2):
    
    if img1.ndim == 3 and img1.shape[2] == 3:
        gray1 = cv2.cvtColor(img1, cv2.COLOR_RGB2GRAY)
        gray2 = cv2.cvtColor(img2, cv2.COLOR_RGB2GRAY)
        return ssim(gray1, gray2)
    return ssim(img1, img2)


def calculate_cd(img1, img2):
    lab1 = rgb2lab(img1 / 255.0)
    lab2 = rgb2lab(img2 / 255.0)
    
    delta_e = np.sqrt(np.sum((lab1 - lab2) ** 2, axis=2))
    return np.mean(delta_e)

def calculate_ie(interpolated, gt):
    return np.mean(np.abs(interpolated.astype(np.float32) - gt.astype(np.float32)))

def interpolate_frames(img0_path, img1_path, output_path, gt_path=None):
    input_tensor, original_size, img0, img1, gt = preprocess_images(img0_path, img1_path, gt_path)
    
    start_time = time.time()
    with torch.no_grad():
        flow, mask, interpolated = model(input_tensor)
    inference_time = time.time() - start_time
    print(f"Inference time: {inference_time:.4f} seconds")
    
    interpolated_img = tensor_to_image(interpolated[0])
    
    interpolated_img = cv2.resize(interpolated_img, (original_size[1], original_size[0]))
    

    interpolated_img_bgr = cv2.cvtColor(interpolated_img, cv2.COLOR_RGB2BGR)
    cv2.imwrite(output_path, interpolated_img_bgr)
    
    metrics = {}
    if gt is not None:
        metrics['psnr'] = calculate_psnr(interpolated_img, gt)
        metrics['ssim'] = calculate_ssim(interpolated_img, gt)
        metrics['cd'] = calculate_cd(interpolated_img, gt)
        metrics['ie'] = calculate_ie(interpolated_img, gt)
        
        print(f"Metrics (compared to ground truth):")
        print(f"  PSNR: {metrics['psnr']:.4f} dB")
        print(f"  SSIM: {metrics['ssim']:.4f}")
        print(f"  Color Difference (CD): {metrics['cd']:.4f}")
        print(f"  Interpolation Error (IE): {metrics['ie']:.4f}")
    
  
    return img0, img1, interpolated_img, gt, metrics


def display_results(img0, img1, interpolated, gt, metrics, output_path):
    
    has_gt = gt is not None
    
    
    plt.figure(figsize=(15, 5 if not has_gt else 10))
    
    
    plt.subplot(2 if has_gt else 1, 3, 1)
    plt.imshow(img0)
    plt.title('Frame 1')
    plt.axis('off')
    
    plt.subplot(2 if has_gt else 1, 3, 2)
    plt.imshow(interpolated)
    plt.title('Interpolated Frame')
    plt.axis('off')
    
    plt.subplot(2 if has_gt else 1, 3, 3)
    plt.imshow(img1)
    plt.title('Frame 2')
    plt.axis('off')
    
    
    if has_gt:
        plt.subplot(2, 3, 4)
        plt.imshow(gt)
        plt.title('Ground Truth')
        plt.axis('off')
        
        plt.subplot(2, 3, 5)
 
        diff = np.abs(interpolated.astype(np.float32) - gt.astype(np.float32))
        plt.imshow(diff.astype(np.uint8))
        plt.title('Difference')
        plt.axis('off')
        
        plt.subplot(2, 3, 6)
        plt.axis('off')
        metrics_text = "\n".join([
            f"PSNR: {metrics['psnr']:.2f} dB",
            f"SSIM: {metrics['ssim']:.4f}",
            f"CD: {metrics['cd']:.2f}",
            f"IE: {metrics['ie']:.2f}"
        ])
        plt.text(0.1, 0.5, metrics_text, fontsize=12)
        plt.title('Metrics')
    
    plt.tight_layout()
    plt.savefig(output_path.replace('.png', '_comparison.png'))
    plt.show()

#CHANGE FILE PATH
test_pairs = [
    
    ("test_frames/frame1.png", "test_frames/frame3.png", "results/scene1_interpolated.png", "test_frames/frame2.png"),
   
]


os.makedirs("results", exist_ok=True)

for test_item in test_pairs:
    img0_path, img1_path, output_path = test_item[0], test_item[1], test_item[2]
    gt_path = test_item[3] if len(test_item) > 3 else None
    
    print(f"Processing: {img0_path} and {img1_path}")
    try:
        img0, img1, interpolated, gt, metrics = interpolate_frames(img0_path, img1_path, output_path, gt_path)
        display_results(img0, img1, interpolated, gt, metrics, output_path)
    except Exception as e:
        print(f"Error processing frames: {e}")
        import traceback
        traceback.print_exc()