code_test/select_threshold.py

import sys
import os
import random
import numpy as np
sys.path.append("/home/shanwei-luo/userdata/mmdetection")
from mmdet.apis import (async_inference_detector, inference_detector,
                        init_detector, show_result_pyplot)
import argparse
#python select_threshold.py --config_file /home/shanwei-luo/teamdata/anomaly_detection_active_learning/model/work_dirs/AD_dsxw_test66_06_10/AD_dsxw_test66_06_10.py --checkpoint_file /home/shanwei-luo/teamdata/anomaly_detection_active_learning/model/work_dirs/AD_dsxw_test66_06_10/latest.pth --images_path /home/shanwei-luo/teamdata/anomaly_detection_active_learning/data0422/smd12_2106_10/test/ --test_batch_size 128
def parse_args():
    parser = argparse.ArgumentParser(description='get best threshold')
    parser.add_argument('--config_file', help='config')
    parser.add_argument('--checkpoint_file', help='checkpoint')
    parser.add_argument('--images_path', help='images')
    parser.add_argument('--test_batch_size', help='images')
    args = parser.parse_args()
    return args

args = parse_args()

class_AOI_name = {"bu_pi_pei":"1","fang_xiang_fan":"2","err.txt_c_not_f":"3", "shang_xi_bu_lia":"4"}

config_file_1 = args.config_file
checkpoint_file_1 = args.checkpoint_file

imgs_ok_path = args.images_path+'ok/'
imgs_ng_path = args.images_path+'ng/'

model_1 = init_detector(config_file_1, checkpoint_file_1, device='cuda:0')

imgs_ok = os.listdir(imgs_ok_path)
imgs_ng = os.listdir(imgs_ng_path)

count_label_ok = len(imgs_ok)
count_label_ng = len(imgs_ng)
print(count_label_ok,count_label_ng)
imgs_labels = []
imgs_name = []
for img in imgs_ok:
    img_name = img.split("@")
    if img_name[2] in class_AOI_name.keys():
        count_label_ok -= 1
        continue
    imgs_labels.append(0)
    imgs_name.append(imgs_ok_path+img)

for img in imgs_ng:
    img_name = img.split("@")
    if img_name[2] in class_AOI_name.keys():
        count_label_ng -= 1
        continue
    imgs_labels.append(1)
    imgs_name.append(imgs_ng_path+img)

print(count_label_ok,count_label_ng, len(imgs_labels))
print("before infer")
index = 0
num = len(imgs_name)
results_1 = []
step = int(args.test_batch_size)
while index<num: 
    index += step
    if index < num:
        results_1_tmp = inference_detector(model_1, imgs_name[index-step:index])
    else:
        results_1_tmp = inference_detector(model_1, imgs_name[index-step:num])
    results_1 += results_1_tmp
    print(len(results_1))
print("after infer")


recall_ok_ng = 0
best_thr = 0.005
for score_thr in np.arange(0.01, 0.3, 0.005):
    imgs_results_1 = []
    for result in results_1:
        res_predict = 0
        for i in result:
            for j in range(i.shape[0]):
                if i[j, 4]>score_thr:
                    res_predict = 1
        imgs_results_1.append(res_predict)

    count_ng = 0
    count_ok = 0
    for i in range(len(imgs_labels)):
        if imgs_labels[i]==0 and imgs_results_1[i]==0:
            count_ok += 1
        if imgs_labels[i]==1 and imgs_results_1[i]==1:
            count_ng += 1
    recall_ok = count_ok/count_label_ok
    recall_ng = count_ng/count_label_ng
    print("score_thr:", score_thr, "  recall(ok):", recall_ok, "  recall(ng):", recall_ng)

    if recall_ok_ng < 0.3*recall_ok + 0.7*recall_ng:
        recall_ok_ng = 0.3*recall_ok + 0.7*recall_ng
        best_thr = score_thr
print("***********************")
print("best threshold:", best_thr)