def write_voc_results_file(all_boxes, test_imgid_list, det_save_dir):
'''
:param all_boxes: is a list. each item reprensent the detections of a img.
the detections is a array. shape is [-1, 7]. [category, score, x, y, w, h, theta]
Note that: if none detections in this img. that the detetions is : []
:param test_imgid_list:
:param det_save_path:
:return:
'''
for cls, cls_id in NAME_LABEL_MAP.items():
if cls == 'back_ground':
continue
print("Writing {} VOC resutls file".format(cls))
tools.mkdir(det_save_dir)
det_save_path = os.path.join(det_save_dir, "det_" + cls + ".txt")
with open(det_save_path, 'wt') as f:
for index, img_name in enumerate(test_imgid_list):
this_img_detections = all_boxes[index]
this_cls_detections = this_img_detections[
this_img_detections[:, 0] == cls_id]
if this_cls_detections.shape[0] == 0:
continue # this cls has none detections in this img
for a_det in this_cls_detections:
f.write('{:s} {:.3f} {:.1f} {:.1f} {:.1f} {:.1f} {:.1f}\n'.
format(img_name, a_det[1], a_det[2], a_det[3],
a_det[4], a_det[5], a_det[6])
) # that is [img_name, score, x, y, w, h, theta]
def do_python_eval(test_imgid_list, test_annotation_path):
AP_list = []
#import matplotlib.pyplot as plt
#import matplotlib.colors as colors
#color_list = colors.cnames.keys()[::6]
for cls, index in NAME_LABEL_MAP.items():
if cls == 'back_ground':
continue
recall, precision, AP = voc_eval(detpath=os.path.join(
cfgs.EVALUATE_DIR, cfgs.VERSION),
test_imgid_list=test_imgid_list,
cls_name=cls,
annopath=test_annotation_path)
AP_list += [AP]
pl.plot(recall,
precision,
lw=2,
label='{} (AP = {:.4f})'
''.format(cls, AP))
print(10 * "__")
pl.xlabel('Recall')
pl.ylabel('Precision')
pl.grid(True)
pl.ylim([0.0, 1.05])
pl.xlim([0.0, 1.0])
pl.title('Precision-Recall')
pl.legend(loc="lower left")
pl.show()
pl.savefig(cfgs.VERSION + '_eval.jpg')
print("hello")
print("mAP is : {}".format(np.mean(AP_list)))
def do_python_eval(test_imgid_list, test_annotation_path):
import matplotlib.colors as colors
import matplotlib.pyplot as plt
AP_list = []
for cls, index in NAME_LABEL_MAP.items():
if cls == 'back_ground':
continue
recall, precision, AP = voc_eval(detpath=cfgs.EVALUATE_R_DIR,
test_imgid_list=test_imgid_list,
cls_name=cls,
annopath=test_annotation_path)
AP_list += [AP]
print("cls : {}|| Recall: {} || Precison: {}|| AP: {}".format(
cls, recall[-1], precision[-1], AP))
# print("{}_ap: {}".format(cls, AP))
# print("{}_recall: {}".format(cls, recall[-1]))
# print("{}_precision: {}".format(cls, precision[-1]))
c = colors.cnames.keys()
c_dark = list(filter(lambda x: x.startswith('dark'), c))
c = ['red', 'orange']
plt.axis([0, 1.2, 0, 1])
plt.plot(recall, precision, color=c_dark[index], label=cls)
plt.legend(loc='upper right')
plt.xlabel('R')
plt.ylabel('P')
plt.savefig('./PR_R.png')
print("mAP is : {}".format(np.mean(AP_list)))
def do_python_eval(test_imgid_list, test_annotation_path):
AP_list = []
# import matplotlib.pyplot as plt
# import matplotlib.colors as colors
# color_list = colors.cnames.keys()[::6]
for cls, index in NAME_LABEL_MAP.items():
if cls == 'back_ground':
continue
recall, precision, AP = voc_eval(detpath=os.path.join(
cfgs.EVALUATE_DIR, cfgs.VERSION),
test_imgid_list=test_imgid_list,
cls_name=cls,
annopath=test_annotation_path,
ovthresh=cfgs.EVAL_THRESHOLD,
use_07_metric=cfgs.USE_07_METRIC)
AP_list += [AP]
print("cls : {}|| Recall: {} || Precison: {}|| AP: {}".format(
cls, recall[-1], precision[-1], AP))
# plt.plot(recall, precision, label=cls, color=color_list[index])
# plt.legend(loc='upper right')
print(10 * "__")
# plt.show()
# plt.savefig(cfgs.VERSION+'.jpg')
print("mAP is : {}".format(np.mean(AP_list)))
def do_python_eval(test_imgid_list, test_annotation_path, iou_thresh=0.5):
import matplotlib.colors as colors
import matplotlib.pyplot as plt
AP_list = []
recall_all = 0
precision_all = 0
cls_to_avg = 0
gt_cls_num_all = 0
tp_all = 0
fp_all = 0
for cls, index in NAME_LABEL_MAP.items():
if cls == 'back_ground':
continue
recall, precision, AP, gt_cls_num, tp, fp = voc_eval(
detpath=cfgs.INFERENCE_SAVE_PATH,
test_imgid_list=test_imgid_list,
cls_name=cls,
annopath=test_annotation_path,
ovthresh=iou_thresh)
# print('cls:', cls, 'gt num:', gt_cls_num)
if np.isnan(AP):
continue
if AP == 0: #recall.size == 0 or precision.size == 0:
if gt_cls_num == 0:
continue
else:
recall, precision, tp, fp = [0], [0], [0], [0]
AP_list += [AP]
recall_all += recall[-1]
precision_all += precision[-1]
cls_to_avg += 1
gt_cls_num_all += gt_cls_num
tp_all += tp[-1]
fp_all += fp[-1]
print(
"cls : {}|| num : {}|| Recall: {} || Precison: {}|| AP: {}".format(
cls, gt_cls_num, recall[-1], precision[-1], AP))
c = colors.cnames.keys()
c_dark = list(filter(lambda x: x.startswith('dark'), c))
c = ['red', 'orange']
plt.axis([0, 1.2, 0.5, 1])
plt.plot(recall[::500],
precision[::500],
color=c_dark[index],
label=cls)
# print(recall, precision)
plt.legend(loc='upper right')
plt.xlabel('R')
plt.ylabel('P')
plt.savefig('./PR_R.png')
print("avg recall is {}".format(recall_all / cls_to_avg))
print("avg precision is {}".format(precision_all / cls_to_avg))
print("avg false alarm is {}".format(fp_all / (tp_all + fp_all)))
print("mAP is : {}".format(np.mean(AP_list)))
def _write_voc_results_file(all_boxes, test_imgid_list, det_save_path):
for cls, cls_ind in NAME_LABEL_MAP.items():
if cls == 'back_ground':
continue
print('Writing {} VOC results file'.format(cls))
with open(det_save_path, 'wt') as f:
for im_ind, index in enumerate(test_imgid_list):
dets = all_boxes[cls_ind][im_ind]
if dets == []:
continue
# the VOCdevkit expects 1-based indices
for k in range(dets.shape[0]):
f.write('{:s} {:.3f} {:.1f} {:.1f} {:.1f} {:.1f}\n'.format(
index, dets[k, -1], dets[k, 0] + 1, dets[k, 1] + 1,
dets[k, 2] + 1, dets[k, 3] + 1))
def do_python_eval(test_imgid_list, test_annotation_path):
import matplotlib.colors as colors
import matplotlib.pyplot as plt
mAP_dict = {}
mPrecision_dict = {}
mRecall_dict = {}
for cls, index in NAME_LABEL_MAP.items():
print(cls)
if cls == 'back_ground':
continue
recall, precision, AP = voc_eval(detpath=cfgs.EVALUATE_H_DIR,
test_imgid_list=test_imgid_list,
cls_name=cls,
annopath=test_annotation_path)
Precision_cls = np.mean(precision)
Recall_cls = np.mean(recall)
print("{}_AP: {}".format(cls, AP))
print("{}_mRecall: {}".format(cls, Recall_cls))
print("{}_mPrecision: {}".format(cls, Precision_cls))
mAP_dict[cls] = AP
mPrecision_dict[cls] = Precision_cls
mRecall_dict[cls] = Recall_cls
c = colors.cnames.keys()
c_dark = list(filter(lambda x: x.startswith('dark'), c))
c = ['blue', 'green']
plt.axis([0, 1.2, 0, 1])
plt.plot(recall, precision, color=c_dark[index], label=cls)
plt.legend(loc='upper right')
plt.xlabel('Recall')
plt.ylabel('Precision')
plt.savefig('./PR_H.png')
print(mAP_dict, mRecall_dict, mPrecision_dict)
total_mAP = np.mean(get_values_from_dict(mAP_dict))
total_mRecall = np.mean(get_values_from_dict(mRecall_dict))
total_mPrecision = np.mean(get_values_from_dict(mPrecision_dict))
print("mAP_H is : {}".format(total_mAP))
print("mRecall_H is : {}".format(total_mRecall))
print("mPrecision_H is : {}".format(total_mPrecision))
# print(mAP, recall, precision)
return total_mAP, total_mRecall, total_mPrecision, mAP_dict, mRecall_dict, mPrecision_dict
def do_python_eval(test_imgid_list, test_annotation_path):
# import matplotlib.colors as colors
# import matplotlib.pyplot as plt
AP_list = []
for cls, index in NAME_LABEL_MAP.items():
if cls == 'back_ground':
continue
recall, precision, AP = voc_eval(detpath=cfgs.EVALUATE_R_DIR,
test_imgid_list=test_imgid_list,
cls_name=cls,
annopath=test_annotation_path,
use_07_metric=cfgs.USE_07_METRIC,
ovthresh=cfgs.EVAL_THRESHOLD)
AP_list += [AP]
print("cls : {}|| Recall: {} || Precison: {}|| AP: {}".format(
cls, recall[-1], precision[-1], AP))
# print("{}_ap: {}".format(cls, AP))
# print("{}_recall: {}".format(cls, recall[-1]))
# print("{}_precision: {}".format(cls, precision[-1]))
r = np.array(recall)
p = np.array(precision)
F1 = 2 * r * p / (r + p)
max_ind = np.argmax(F1)
print('F1:{} P:{} R:{}'.format(F1[max_ind], p[max_ind], r[max_ind]))
# c = colors.cnames.keys()
# c_dark = list(filter(lambda x: x.startswith('dark'), c))
# c = ['red', 'orange']
# plt.axis([0, 1.2, 0, 1])
# plt.plot(recall, precision, color=c_dark[index], label=cls)
# plt.legend(loc='upper right')
# plt.xlabel('R')
# plt.ylabel('P')
# plt.savefig('./PR_R.png')
print("mAP is : {}".format(np.mean(AP_list)))