def test_eval(model_obj, data_dict, num_anomaly_sets):
test_X = data_dict['test']
test_scores = model_obj.score_samples(test_X)
auc_list = []
for idx in range(num_anomaly_sets):
key = 'anom_' + str(idx + 1)
anom_X = data_dict[key]
anom_scores = model_obj.score_samples(anom_X)
auPR = eval.eval(anom_scores, test_scores, order='descending')
auc_list.append(auPR)
print("AUC : {:0.4f} ".format(auPR))
_mean = np.mean(auc_list)
_std = np.std(auc_list)
print(' Mean AUC ', np.mean(auc_list))
print(' AUC std', np.std(auc_list))
return _mean, _std
def main(opt):
logger.info('* batch_size: %d' % opt.batch_size)
logger.info('* model_path: {}'.format(opt.models[0]))
testset = loads_json(opt.data, 'load test file')
translator = build_translator(opt, report_score=False)
translator.model.cpu()
translated = translator.translate(data_path=opt.data,
batch_size=opt.batch_size)
# find first 3 not similar distractors
hypothesis = {}
for translation in translated:
question_id = str(translation.ex_raw.id['file_id']) + '_' + str(
translation.ex_raw.id['question_id'])
pred1 = translation.pred_sents[0]
pred2, pred3 = None, None
for pred in translation.pred_sents[1:]:
if jaccard_similarity(pred1, pred) < 0.5:
if pred2 is None:
pred2 = pred
else:
if pred3 is None:
if jaccard_similarity(pred2, pred) < 0.5:
pred3 = pred
if pred2 is not None and pred3 is not None:
break
if pred2 is None:
pred2 = translation.pred_sents[1]
if pred3 is None:
pred3 = translation.pred_sents[2]
else:
if pred3 is None:
pred3 = translation.pred_sents[1]
hypothesis[question_id] = [pred1, pred2, pred3]
reference = {}
for sample in testset:
question_id = str(sample['id']['file_id']) + '_' + str(
sample['id']['question_id'])
if question_id not in reference.keys():
reference[question_id] = [sample['distractor']]
else:
reference[question_id].append(sample['distractor'])
_ = eval(hypothesis, reference)
def main(args):
ea = EA(1)
ea.load_config(args.config)
reseedPeriod = int(args.reseed)
taskNum = int(args.task_num)
np.random.seed(0)
seed = np.random.randint(0, 2**32 - 1, size=(taskNum), dtype=np.uint32)
seed = seed.tolist()
print(seed)
for i in range(int(args.generation)):
if ((reseedPeriod > 0) and (i % reseedPeriod == 0)):
for j in range(taskNum):
seed[j] = random.randint(0, 2**32 - 1)
ea_time = time.time()
pop = ea.ask()
ea_time = time.time() - ea_time
fitnesses = []
workloads = []
num_workers = int(args.num_workers) - 1
gc.collect()
prep_time = time.time()
for j in range(len(pop)):
workloads.append((pop[j], args.task, seed))
prep_time = time.time() - prep_time
eval_time = time.time()
if (num_workers > 1):
with mp.Pool(num_workers) as pool:
results = pool.map(eval, workloads)
else:
results = []
for w in workloads:
results.append(eval(w))
eval_time = time.time() - eval_time
ea.tell(results, args.task, seed)
ea.write_history(args.output)
print(
'iter: {0} fit: {1}, Q: {2}, ea_time: {3}, prep_time: {4}, eval_time: {5}, max_depth:{6}'
.format(i, ea.fitnesses[0], np.mean(ea.Q), ea_time, prep_time,
eval_time, ea.pop[0].maxDepth))
boxes, probs = test_img(net, data_encoder, img, scale, thresh=0.5)
# print(boxes)
face_num = len(boxes) / 4
s = str(int(face_num)) + " "
for i in range(int(face_num)):
x1 = int(boxes[i * 4] * 1)
y1 = int(boxes[i * 4 + 1] * 1)
x2 = int(boxes[i * 4 + 2] * 1)
y2 = int(boxes[i * 4 + 3] * 1)
s = s + str(x1) + " " + str(y1) + " " + str(x2) + " " + str(y2) + " "
print(file, s[:-1])
# cv2.imwrite("./all_test_detected/" + file + "_result.jpg", img)
#test img
# if 0:
# files = os.listdir('./eval/badcase_edge/')
# for file in files:
# imgPath = os.path.join('./eval/badcase_edge/',file)
# print(imgPath)
# t = time.time()
# for k in range(30):
# boxes,probs = test_img(net,data_encoder,imgPath,scale,thresh=0.5)
# print( 'each time: ',(time.time()-t)/30.0)
if __name__ == '__main__':
test_faceInCar()
eval(label_path, outfile, output_recall_far_file)
#test()
def main(opt):
testset = loads_json(opt.data, 'load test file')
f = open("data/corpus.json", "r")
data = f.readlines()
samples = []
for i in data:
samples.append(json.loads(i)["sents"])
corpus = []
for i in samples:
corpus.append(" ".join([" ".join(sent) for sent in i]))
vectorizer = TfidfVectorizer()
tfidf = vectorizer.fit(corpus)
translator = build_translator(opt, report_score=False)
translated = translator.translate(data_path=opt.data,
batch_size=opt.batch_size,
tfidf=tfidf)
# find first 3 not similar distractors
hypothesis = {}
for translation in translated:
sample_id = str(translation.ex_raw.id).split("_")
question_id = sample_id[0] + "_" + sample_id[1]
pred1 = translation.pred_sents[0]
pred2, pred3 = None, None
for pred in translation.pred_sents[1:]:
if jaccard_similarity(pred1, pred) < 0.5:
if pred2 is None:
pred2 = pred
else:
if pred3 is None:
if jaccard_similarity(pred2, pred) < 0.5:
pred3 = pred
if pred2 is not None and pred3 is not None:
break
if pred2 is None:
pred2 = translation.pred_sents[1]
if pred3 is None:
pred3 = translation.pred_sents[2]
else:
if pred3 is None:
pred3 = translation.pred_sents[1]
hypothesis[question_id] = [pred1, pred2, pred3]
torch.save(hypothesis, "translated.hypothesis.pt")
reference = {}
f = open(opt.output, "w")
for sample in testset:
line = copy.deepcopy(sample)
sample_id = sample["id"].split("_")
question_id = sample_id[0] + "_" + sample_id[1]
if question_id not in reference.keys():
reference[question_id] = [sample['distractor']]
else:
reference[question_id].append(sample['distractor'])
line["pred"] = hypothesis[question_id]
f.write(json.dumps(line) + "\n")
f.close()
_ = eval(hypothesis, reference)