def analyse(api_key, endpoint, doi, project_id, result_id, project_short_name,
path, throttle, **kwargs):
"""Analyse Convert-a-Card results."""
e = enki.Enki(api_key, endpoint, project_short_name, all=1)
result = enki.pbclient.find_results(project_id,
id=result_id,
limit=1,
all=1)[0]
df = helpers.get_task_run_df(e, result.task_id)
df = df.loc[:, df.columns.isin(VALID_KEYS)]
df = helpers.drop_empty_rows(df)
n_task_runs = len(df.index)
# Initialise the result
defaults = {k: "" for k in df.keys()}
result.info = helpers.init_result_info(doi, path, defaults)
has_answers = not df.empty
has_matches = helpers.has_n_matches(df, n_task_runs, MATCH_PERCENTAGE)
# Matching answers
if has_answers and has_matches:
for k in df.keys():
result.info[k] = df[k].value_counts().idxmax()
# Varied answers
elif has_answers:
result.info['analysis_complete'] = False
enki.pbclient.update_result(result)
time.sleep(throttle)
def download_task_runs():
with open(KEY_PATH, "r") as f:
key = f.read()
e = enki.Enki(api_key=key,
endpoint='http://localhost:5000',
project_short_name='opentaal',
all=1)
e.get_all()
return e.task_runs_df
def test_basic(self):
"""Test basic method works."""
with patch('enki.Enki', autospec=True):
enki_mock = enki.Enki(endpoint='server',
api_key='api',
project_short_name='project')
enki_mock.tasks = []
res = basic(**self.payload)
assert enki_mock.get_tasks.called
assert enki_mock.get_task_runs.called
assert res == 'OK', res
def analyse_all(**kwargs):
"""Analyse all Convert-a-Card results."""
print kwargs
e = enki.Enki(kwargs['api_key'],
kwargs['endpoint'],
kwargs['project_short_name'],
all=1)
results = object_loader.load(enki.pbclient.find_results,
project_id=e.project.id,
all=1)
for result in results:
kwargs['project_id'] = e.project.id
kwargs['result_id'] = result.id
print kwargs
analyse(**kwargs.copy())
def analyse_all_selections(**kwargs):
"""Analyse all In the Spotlight selection results."""
e = enki.Enki(kwargs['api_key'], kwargs['endpoint'],
kwargs['project_short_name'], all=1)
results = object_loader.load(enki.pbclient.find_results,
project_id=e.project.id, all=1)
for result in results:
kwargs['project_id'] = e.project.id
kwargs['result_id'] = result.id
analyse_selections(**kwargs.copy())
helpers.send_mail({
'recipients': kwargs['mail_recipients'],
'subject': 'Analysis complete',
'body': '''
All {0} results for {1} have been analysed.
'''.format(len(results), e.project.name)
})
def analyse_selections(api_key, endpoint, project_id, result_id, path, doi,
project_short_name, throttle, **kwargs):
"""Analyse In the Spotlight selection results."""
e = enki.Enki(api_key, endpoint, project_short_name, all=1)
result = enki.pbclient.find_results(project_id, id=result_id, limit=1,
all=1)[0]
df = helpers.get_task_run_df(e, result.task_id)
# Flatten annotations into a single list
anno_list = df['info'].tolist()
anno_list = list(itertools.chain.from_iterable(anno_list))
defaults = {'annotations': []}
result.info = helpers.init_result_info(doi, path, defaults)
clusters = []
comments = []
# Cluster similar regions
for anno in anno_list:
if anno['motivation'] == 'commenting':
comments.append(anno)
continue
elif anno['motivation'] == 'tagging':
r1 = get_rect_from_selection(anno)
matched = False
for cluster in clusters:
r2 = get_rect_from_selection(cluster)
overlap_ratio = get_overlap_ratio(r1, r2)
if overlap_ratio > MERGE_RATIO:
matched = True
r3 = merge_rects(r1, r2)
update_selector(cluster, r3)
if not matched:
update_selector(anno, r1) # still update to round rect params
clusters.append(anno)
else: # pragma: no cover
raise ValueError('Unhandled motivation')
result.info['annotations'] = clusters + comments
enki.pbclient.update_result(result)
time.sleep(throttle)
def basic(**kwargs):
"""A basic analyzer."""
e = enki.Enki(endpoint=settings.endpoint,
api_key=settings.api_key,
project_short_name=kwargs['project_short_name'])
e.get_tasks(task_id=kwargs['task_id'])
e.get_task_runs()
for t in e.tasks: # pragma: no cover
desc = e.task_runs_df[t.id]['info'].describe()
# print "The top answer for task.id %s is %s" % (t.id, desc['top'])
value_counts = e.task_runs_df[t.id]['info'].value_counts()
analysis = dict(value_counts)
summary = dict(desc)
result = enki.pbclient.find_results(project_id=kwargs['project_id'],
id=kwargs['result_id'])[0]
result.info = dict(summary=summary, analysis=analysis)
enki.pbclient.update_result(result)
with open('./static/results.json', 'w') as f:
f.write(json.dumps(kwargs))
return "OK"
def relevance(**kwargs):
headers = {
'Content-Type': 'application/json',
}
# Obtain task run results from Pybossa:
e = enki.Enki(api_key=settings.api_key, endpoint=settings.endpoint, project_short_name=kwargs['project_short_name'], all=1)
e.get_tasks(task_id=kwargs['task_id'])
post_id = e.tasks_df['_id'][kwargs['task_id']]
e.get_task_runs()
df = e.task_runs_df[kwargs['task_id']]
# Calculate total number of answers for the task and create a subset DF:
total = len(df[df.columns[0]])
subset = df[df.columns[0]].apply(pd.Series)[0].apply(pd.Series)
# Count the number of times the image was considered relevant:
trueCount = len(subset[subset[subset.columns[0]]==True])
# Calculate relevance:
relevance = trueCount/total
#Prepare payload:
payload = {
"name":"relevance",
"post":post_id,
"value":str(relevance)
}
# e2mc data API credentials:
username = # e2mc data api username. Although harcoded here, it could be easily included in the settings file
password = # e2mc data api password. Although harcoded here, it could be easily included in the settings file
# Push payload into data API
r = requests.post('http://131.175.120.92:5555/e2mc/datapi/v1.1/tags', auth=(username, password), headers=headers, data=json.dumps(payload))
# Create record file:
with open('./static/results.json', 'w') as f:
f.write(json.dumps(kwargs))
return r
def basic(**kwargs):
"""A basic analyzer."""
e = enki.Enki(endpoint=settings.endpoint,
api_key=settings.api_key,
project_short_name=kwargs['project_short_name'],
all=1)
if kwargs['task_id'] != 95049:
e.get_tasks(task_id=kwargs['task_id'])
e.get_task_runs()
labels = ['task_run_id', 'speciesID', 'speciesScientificName',
'speciesCommonName', 'animalCount']
for t in e.tasks:
data = []
project_id = t.project_id
for tr in e.task_runs[t.id]:
for datum in tr.info['answer']:
data.append(datum)
df = pd.DataFrame(data)
# If 5 first answers is nan (nothing here) mark task
# as completed
vc = get_count_nan(df)
print vc.index[0]
print vc.values[0]
if len(e.task_runs[t.id]) == 5:
msg = "The five taskruns reported no animal"
if type(vc) == pd.Series and ((str(vc.index[0]) == 'nan' or
vc.index[0] == -1) and vc.values[0] == 5):
result = enki.pbclient.find_results(project_id=kwargs['project_id'],
id=kwargs['result_id'],all=1)
if len(result) > 0:
return create_result(t, settings.no_animal, result[0])
else:
task = get_task(t.project_id, t.id)
task.n_answers += 1
task.state = 'ongoing'
return enki.pbclient.update_task(task)
else:
if (str(vc.index[0]) == 'nan' or vc.index[0] == -1) and vc.values[0] >= 10:
msg = "10 taskruns reported no animal"
result = enki.pbclient.find_results(project_id=kwargs['project_id'],
id=kwargs['result_id'],all=1)
if len(result) > 0:
return create_result(t, settings.no_animal, result[0])
else:
answers = get_consensus(df, th=10)
if len(answers) == 0:
if len(e.task_runs[t.id]) < 25:
msg = "No consensus. Asking for one more answer."
task = get_task(t.project_id, t.id)
task.n_answers += 1
task.state = 'ongoing'
return enki.pbclient.update_task(task)
else:
result = enki.pbclient.find_results(project_id=kwargs['project_id'],
id=kwargs['result_id'],all=1)
if len(result) > 0:
return create_result(t, settings.no_consensus,
result[0])
else:
for a in answers:
iucn_red_list_status, species = get_red_list_status(a['speciesScientificName'], project_id)
a['speciesCommonName'] = species
a['iucn_red_list_status'] = iucn_red_list_status
a['imageURL'] = t.info.get('image', None)
a['deploymentID'] = t.info.get('deploymentID', None)
a['deploymentLocationID'] = t.info.get('deploymentLocationID', None)
a['Create_time'] = t.info.get('Create_time')
result = enki.pbclient.find_results(project_id=kwargs['project_id'],
id=kwargs['result_id'],all=1)
if len(result) > 0:
result = result[0]
if len(answers) == 1:
result.info = answers[0]
if len(answers) >= 2:
result.info = dict(answers=answers)
give_badges(e, t, answers, result)
result = enki.pbclient.update_result(result)
return 'OK'
return "OK"