epidermal/webapp_model.py at main · szerrade-uw/epidermal · 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
147
148
149
#!/usr/bin/env python
# Model overview and functions

import os
from flask import render_template, Blueprint, redirect, send_from_directory
import db
from bson.objectid import ObjectId
import re

from webapp_base import pop_last_error, set_notice
from webapp_admin import requires_admin
from config import config


bp_model = Blueprint('model', __name__, template_folder='templates')


def enqueue_images_for_model(model_id):
    undeleted_dataset_ids = set([ds['_id'] for ds in db.get_datasets()])
    # Find all images not processed by the given model
    all_sample_ids = set([s['_id'] for s in db.get_samples() if s['dataset_id'] in undeleted_dataset_ids])
    machine_annotations = set([a['sample_id'] for a in db.get_machine_annotations_for_model(model_id=model_id)])
    enqueued_samples = set([s['sample_id'] for s in db.get_queued_samples(model_id=model_id) if 'sample_id' in s])
    sample_ids_to_enqueue = all_sample_ids - machine_annotations - enqueued_samples
    for sample_id in sample_ids_to_enqueue:
        db.queue_sample(sample_id=sample_id, model_id=model_id)
    return len(sample_ids_to_enqueue)


@bp_model.route('/enqueue_images/<model_id_s>')
#@requires_admin
def enqueue_images(model_id_s):
    model_id = ObjectId(model_id_s)
    n = enqueue_images_for_model(model_id)
    set_notice('Enqueued %d images.' % n)
    return redirect('/model/' + model_id_s)


@bp_model.route('/set_primary_model/<model_id_s>')
#@requires_admin
def set_primary_model(model_id_s):
    model_id = ObjectId(model_id_s)
    db.set_primary_model(model_id)
    set_notice('Updated primary model.')
    return redirect('/model/' + model_id_s)


@bp_model.route('/enqueue_all_images')
#@requires_admin
def enqueue_all_images():
    n = 0
    m = 0
    for model in db.get_models(details=False, status=db.model_status_trained):
        if not model['primary']:
            nm = enqueue_images_for_model(model['_id'])
            if nm:
                n += nm
                m += 1
    set_notice('Enqueued %d images from %d models.' % (n, m))
    return redirect('/admin')


def enqueue_validation_sets_for_model(train_model_id):
    existing_results = db.get_all_validation_results(train_model_id=train_model_id)
    processed_val_model_ids = set([r['validation_model_id'] for r in existing_results if r['image_subset'] == 'train'])
    n = 0
    for validation_model in db.get_models(details=False, status=db.model_status_dataset)\
            + db.get_models(details=False, status=db.model_status_trained):
        if validation_model['_id'] not in processed_val_model_ids:
            db.queue_validation(train_model_id=train_model_id, validation_model_id=validation_model['_id'])
            n += 1
    return n


@bp_model.route('/enqueue_validation_sets/<model_id_s>')
#@requires_admin
def enqueue_validation_sets(model_id_s):
    model_id = ObjectId(model_id_s)
    n = enqueue_validation_sets_for_model(model_id)
    set_notice('Enqueued %d sets.' % n)
    return redirect('/model/' + model_id_s)


@bp_model.route('/enqueue_all_validation_sets')
#@requires_admin
def enqueue_all_validation_sets():
    n = 0
    m = 0
    for model in db.get_models(details=False, status=db.model_status_trained):
        nm = enqueue_validation_sets_for_model(model['_id'])
        if nm:
            n += nm
            m += 1
    set_notice('Enqueued %d sets from %d models.' % (n, m))
    return redirect('/admin')


@bp_model.route('/model/<model_id_s>')
#@requires_admin
def model_page(model_id_s):
    model_id = ObjectId(model_id_s)
    model = dict(db.get_model_by_id(model_id))
    model_data = [(k, v) for k, v in sorted(model.items())]
    exec_base_path = os.path.abspath(os.path.join(os.path.dirname(__file__), 'cnn'))
    exec_path = os.path.join(exec_base_path, str(model_id))
    log_filename = os.path.join(exec_path, 'train.log')
    if os.path.isfile(log_filename):
        log_data = open(log_filename, 'rt').read()
    else:
        log_data = 'logfile not found: %s' % log_filename
    # Get validation results by model
    validation_results = db.get_all_validation_results(train_model_id=model_id)
    for vr in validation_results:
        vr['image_set'] = '(%s) %s' % (vr['image_subset'], db.get_model_by_id(vr['validation_model_id'])['name'])
        cm = vr['confusion_matrix']
        vr['accuracy'] = '%.1f' % (100.0 * (cm[0][0] + cm[1][1]) / (cm[0][0] + cm[1][1] + cm[1][0] + cm[0][1]))
    validation_results = sorted(validation_results, key=lambda f: f['image_set'])
    # Get validation results by dataset
    validation_results_ds = db.get_all_validation_results(validation_model_id=model_id)
    for vr in validation_results_ds:
        vr['model_name'] = '(%s) %s' % (vr['image_subset'], db.get_model_by_id(vr['train_model_id'])['name'])
        cm = vr['confusion_matrix']
        vr['accuracy'] = '%.1f' % (100.0 * (cm[0][0] + cm[1][1]) / (cm[0][0] + cm[1][1] + cm[1][0] + cm[0][1]))
    validation_results_ds = sorted(validation_results_ds, key=lambda f: f['model_name'])
    return render_template('model.html', model_id=model_id_s, name=model['name'], model_data=model_data, log=log_data,
                           validation_results=validation_results, validation_results_ds=validation_results_ds,
                           error=pop_last_error())


@bp_model.route('/validation/<val_id_s>')
#@requires_admin
def validation_page(val_id_s):
    val_id = ObjectId(val_id_s)
    val = dict(db.get_validation_results_by_id(val_id))
    train_model = db.get_model_by_id(val['train_model_id'])
    val_model = db.get_model_by_id(val['validation_model_id'])
    train_name = train_model['name']
    val_name = '%s (%s)' % (val_model['name'], val['image_subset'])
    worst_predictions = val['worst_predictions']
    return render_template('validation.html', val_id=val['validation_model_id'], train_name=train_name,
                           val_name=val_name, worst_predictions=worst_predictions, error=pop_last_error())


@bp_model.route('/image_sample/<train_id_s>/<path:sample_name>')
#@requires_admin
def image_sample(train_id_s, sample_name):
    subpath = os.path.join(train_id_s, 'samples', sample_name)
    print('GETTING', subpath)
    return send_from_directory(config.get_train_data_path(), subpath)