SPDET/spdet.py at master · zcq15/SPDET · 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
from tqdm import tqdm
from models.options import Options
Options().paser()
from models.gargs import _args
import torch
from torch.utils.data import DataLoader

from models.network import Network
from datasets.pnvsdepth import PNVSDepth,EasyPNVSDepth,HardPNVSDepth
from datasets.threed60 import ThreeD60,ThreeD60M3D,ThreeD60S2D3D
# from datasets.panosuncg import PanoSUNCG

from models.metrics.metrics import Photometric,Perceptual,UVEdge,UVSmooth,BerHu,Smooth
from models.metrics.metrics_eval_perspective import L1,SqRel,RMSE,Log10RMSE,AbsRel,Delta
from models.metrics.metrics_eval_spherical import SPL1,SPAbsRel,SPSqRel,SPRMSE,SPLogERMSE,SPDelta

import os
from collections import OrderedDict
import glob

def run(funcs,weights,warmup,metrics):
    model = Network(funcs)
    dataloader_train,dataloader_test,sampler_train,sampler_test = run_init()
    test = run_exec(model,
            dataloader_train,dataloader_test,
            sampler_train,sampler_test,
            weights,warmup,metrics)
    return test

def run_init():

    dataset_dict = {
        'PNVSDepth':PNVSDepth,
        'EasyPNVSDepth':EasyPNVSDepth,
        'HardPNVSDepth':HardPNVSDepth,
        '3D60':ThreeD60,
        '3D60M3D':ThreeD60M3D,
        '3D60S2D3D':ThreeD60S2D3D,
        # 'PanoSUNCG':PanoSUNCG
    }

    dataset = dataset_dict[_args['arch']['dataset']]

    dataset_train = dataset(_args['data']['datadir'],area='train',imgsize=_args['data']['imgsize'],local_rank=_args['arch']['rank'])
    sampler_train = torch.utils.data.distributed.DistributedSampler(dataset_train)
    dataloader_train = DataLoader(dataset_train,_args['data']['batch_size'],num_workers=_args['data']['threads'],\
        sampler=sampler_train,pin_memory=True,drop_last=_args['data']['droplast'])

    dataset_test = dataset(_args['data']['datadir'],area='test',imgsize=_args['data']['imgsize'],local_rank=_args['arch']['rank'])
    sampler_test = torch.utils.data.distributed.DistributedSampler(dataset_test)
    dataloader_test = DataLoader(dataset_test,_args['data']['batch_size'],num_workers=_args['data']['threads'],sampler=sampler_test,pin_memory=True,drop_last=_args['data']['droplast'])

    return dataloader_train,dataloader_test,sampler_train,sampler_test

def run_exec(model,
             dataloader_train,dataloader_test,
             sampler_train,sampler_test,
             weights,warmup,metrics):

    volumes_test = OrderedDict({'Size':0})
    for key in metrics:
        volumes_test[key] = 0
    if _args['arch']['mode'] == 'train':
        start = model.epoch
        end = _args['optim']['epochs']
        iters = range(start,end)
        for epoch in iters:

            _args['arch']['mode'] = 'train'
            sampler_train.set_epoch(model.epoch)
            model.train(dataloader_train,weights,warmup)

            _args['arch']['mode'] = 'test'
            sampler_test.set_epoch(model.epoch)
            model.test(dataloader_test,metrics)
            _args['arch']['mode'] = 'train'

    else:
        sampler_test.set_epoch(model.epoch)
        model.test(dataloader_test,metrics)

    del model,dataloader_train,dataloader_test
    torch.cuda.empty_cache()
    return volumes_test

if __name__ == '__main__':

    if _args['optim']['supervise'] == 'scatter':

        weights = _args['optim']['weights'] or {'P':1.0,'V':0.01,'E':0.01,'S':0.01}
        warmup = {}
        metrics =['WAbs','WSq','WMAE','WRMSE','WLogERMSE','WD1','WD2','WD3']

        funcs = OrderedDict({
            'P':Photometric(branch='tar_rgb_scatter',target='target',valid='tar_valid_scatter'),
            'V':Perceptual(branch='tar_rgb_scatter',target='target',valid='tar_valid_scatter'),
            'E':UVEdge(branch='tar_uv_scatter',target='source'),
            'S':UVSmooth(branch='tar_uv_scatter',target='source'),

            'MAE':L1(max=_args['data']['max']),'RMSE':RMSE(max=_args['data']['max']),'Log10RMSE':Log10RMSE(max=_args['data']['max']),
            'Abs':AbsRel(max=_args['data']['max']),
            'D1':Delta(th=1.25,max=_args['data']['max']),'D2':Delta(th=1.25**2,max=_args['data']['max']),'D3':Delta(th=1.25**3,max=_args['data']['max']),

            'WMAE':SPL1(max=_args['data']['max']),'WRMSE':SPRMSE(max=_args['data']['max']),'WLogERMSE':SPLogERMSE(max=_args['data']['max']),
            'WAbs':SPAbsRel(max=_args['data']['max']),'WSq':SPSqRel(max=_args['data']['max']),
            'WD1':SPDelta(th=1.25,max=_args['data']['max']),'WD2':SPDelta(th=1.25**2,max=_args['data']['max']),'WD3':SPDelta(th=1.25**3,max=_args['data']['max'])

        })

    elif _args['optim']['supervise'] == 'inter':

        weights = _args['optim']['weights'] or {'P':1.0,'V':0.01,'S':0.01}
        warmup = {}
        metrics =['WAbs','WSq','WMAE','WRMSE','WLogERMSE','WD1','WD2','WD3']

        funcs = OrderedDict({
            'P':Photometric(branch='src_rgb_inter',target='source'),
            'V':Perceptual(branch='src_rgb_inter',target='source'),
            'S':Smooth(branch='src_depth',target='source'),

            'MAE':L1(max=_args['data']['max']),'RMSE':RMSE(max=_args['data']['max']),'Log10RMSE':Log10RMSE(max=_args['data']['max']),
            'Abs':AbsRel(max=_args['data']['max']),
            'D1':Delta(th=1.25,max=_args['data']['max']),'D2':Delta(th=1.25**2,max=_args['data']['max']),'D3':Delta(th=1.25**3,max=_args['data']['max']),

            'WMAE':SPL1(max=_args['data']['max']),'WRMSE':SPRMSE(max=_args['data']['max']),'WLogERMSE':SPLogERMSE(max=_args['data']['max']),
            'WAbs':SPAbsRel(max=_args['data']['max']),'WSq':SPSqRel(max=_args['data']['max']),
            'WD1':SPDelta(th=1.25,max=_args['data']['max']),'WD2':SPDelta(th=1.25**2,max=_args['data']['max']),'WD3':SPDelta(th=1.25**3,max=_args['data']['max'])

        })

    else:

        weights = _args['optim']['weights'] or {'B':1.0}
        warmup = {}
        metrics =['Abs','SqRel','MAE','RMSE','Log10RMSE','D1','D2','D3']

        funcs = OrderedDict({
            'B':BerHu(max=_args['data']['max']),'L1':L1(max=_args['data']['max']),

            'MAE':L1(max=_args['data']['max']),'RMSE':RMSE(max=_args['data']['max']),'Log10RMSE':Log10RMSE(max=_args['data']['max']),'Abs':AbsRel(max=_args['data']['max']), 'SqRel':SqRel(max=_args['data']['max']),
            'D1':Delta(th=1.25,max=_args['data']['max']),'D2':Delta(th=1.25**2,max=_args['data']['max']),'D3':Delta(th=1.25**3,max=_args['data']['max'])
        })

    run(funcs,weights,warmup,metrics)