layered-neural-atlases/only_evaluate.py at main · ykasten/layered-neural-atlases · 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
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
from implicit_neural_networks import IMLP
from unwrap_utils import load_input_data
import time
import torch
import numpy as np
import sys

import argparse
from evaluate import evaluate_model
import os

import json

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
from pathlib import Path


def main(training_folder, frames_folder, mask_rcnn_folder,  output_folder, video_name,runinng_command):
    # read config:
    with open("%s/config.json" % training_folder) as f:
        config = json.load(f)

    maximum_number_of_frames = config["maximum_number_of_frames"]
    resx = np.int64(config["resx"])
    resy = np.int64(config["resy"])

    positional_encoding_num_alpha = config["positional_encoding_num_alpha"]

    number_of_channels_atlas = config["number_of_channels_atlas"]
    number_of_layers_atlas = config["number_of_layers_atlas"]

    number_of_channels_alpha = config["number_of_channels_alpha"]
    number_of_layers_alpha = config["number_of_layers_alpha"]
    uv_mapping_scale = config["uv_mapping_scale"]
    use_positional_encoding_mapping1 = config["use_positional_encoding_mapping1"]
    number_of_positional_encoding_mapping1 = config["number_of_positional_encoding_mapping1"]
    number_of_layers_mapping1 = config["number_of_layers_mapping1"]
    number_of_channels_mapping1 = config["number_of_channels_mapping1"]

    use_positional_encoding_mapping2 = config["use_positional_encoding_mapping2"]
    number_of_positional_encoding_mapping2 = config["number_of_positional_encoding_mapping2"]
    number_of_layers_mapping2 = config["number_of_layers_mapping2"]
    number_of_channels_mapping2 = config["number_of_channels_mapping2"]
    derivative_amount = config["derivative_amount"]
    data_folder = Path(config["data_folder"])
    vid_name = data_folder.name
    vid_root = data_folder.parent
    optical_flows_mask, video_frames, _, mask_frames, _, _, _, optical_flows = load_input_data(
            resy, resx, maximum_number_of_frames, data_folder, True, True, vid_root, vid_name)
    number_of_frames = video_frames.shape[3]

    # load networks' weights:
    model_F_mapping1 = IMLP(
        input_dim=3,
        output_dim=2,
        hidden_dim=number_of_channels_mapping1,
        use_positional=use_positional_encoding_mapping1,
        positional_dim=number_of_positional_encoding_mapping1,
        num_layers=number_of_layers_mapping1,
        skip_layers=[]).to(device)
    model_F_mapping2 = IMLP(
        input_dim=3,
        output_dim=2,
        hidden_dim=number_of_channels_mapping2,
        use_positional=use_positional_encoding_mapping2,
        positional_dim=number_of_positional_encoding_mapping2,
        num_layers=number_of_layers_mapping2,
        skip_layers=[]).to(device)

    model_F_atlas = IMLP(
        input_dim=2,
        output_dim=3,
        hidden_dim=number_of_channels_atlas,
        use_positional=True,
        positional_dim=10,
        num_layers=number_of_layers_atlas,
        skip_layers=[4, 7]).to(device)

    model_alpha = IMLP(
        input_dim=3,
        output_dim=1,
        hidden_dim=number_of_channels_alpha,
        use_positional=True,
        positional_dim=positional_encoding_num_alpha,
        num_layers=number_of_layers_alpha,
        skip_layers=[]).to(device)

    checkpoint = torch.load("%s/checkpoint" % training_folder)

    model_F_atlas.load_state_dict(checkpoint["F_atlas_state_dict"])
    model_F_atlas.eval()
    model_F_atlas.to(device)

    model_F_mapping1.load_state_dict(checkpoint["model_F_mapping1_state_dict"])
    model_F_mapping1.eval()
    model_F_mapping1.to(device)

    model_F_mapping2.load_state_dict(checkpoint["model_F_mapping2_state_dict"])
    model_F_mapping2.eval()
    model_F_mapping2.to(device)

    model_alpha.load_state_dict(checkpoint["model_F_alpha_state_dict"])
    model_alpha.eval()
    model_alpha.to(device)

    folder_time = time.time()


    Path(os.path.join(output_folder, '%s_%06d' % (video_name, folder_time))).mkdir(parents=True, exist_ok=True)
    file1 = open("%s/runinng_command" % os.path.join(output_folder, '%s_%06d' % (video_name, folder_time)), "w")
    file1.write(runinng_command)
    file1.close()
    start_iteration = checkpoint["iteration"]
    # run evaluation:
    evaluate_model(model_F_atlas,resx,resy,number_of_frames,model_F_mapping1,model_F_mapping2,model_alpha,video_frames,os.path.join(output_folder, '%s_%06d' % (video_name, folder_time)),start_iteration,mask_frames,0,0,video_name,derivative_amount,uv_mapping_scale,optical_flows,optical_flows_mask,device,save_checkpoint=False,show_atlas_alpha=True)

if __name__ == "__main__":
    parser = argparse.ArgumentParser(description='Process some integers.')

    parser.add_argument('--trained_model_folder', type=str,
                        help='the folder that contains the trained model')


    parser.add_argument('--data_folder', type=str,
                        help='the folder that contains the masks produced by Mask-RCNN and the images ')
    parser.add_argument('--video_name', type=str,
                        help='the name of the video that should be evaluated')

    parser.add_argument('--output_folder', type=str,
                        help='the folder that will contains the output evaluation ')


    args = parser.parse_args()

    training_folder = args.trained_model_folder
    video_name = args.video_name

    frames_folder = os.path.join(args.data_folder, video_name)
    mask_rcnn_folder = os.path.join(args.data_folder, video_name) + "_maskrcnn"
    output_folder = args.output_folder
    Path(output_folder).mkdir(parents=True, exist_ok=True)


    main(training_folder, frames_folder, mask_rcnn_folder, output_folder, video_name, ' '.join(sys.argv))