update: label map

.gitignore

@@ -0,0 +1,6 @@
+*.pyc
+__pycache__/
+__pycache__/*
+.idea/
+.DS_Store
+results/*

evaluation.py

@@ -38,11 +38,11 @@ def load_all_files(self):
             print('Incorrect threshold value! It should be in [0, 1]. Please check and retry ~')
             return 0
         pre_files = [x for x in os.listdir(prediction_path) if (x.endswith('txt') or x.endswith('xml'))]
-        print ("Num of prediction files: ", len(pre_files))
+        print("Num of prediction files: ", len(pre_files))
         gt_files = os.listdir(gt_path)
-        print ("Num of ground truth files: ", len(gt_files))
+        print("Num of ground truth files: ", len(gt_files))
         if len(pre_files) != len(gt_files):
-            print("groundtruths' size does not match predictions' size， please check ~ ")
+            print("ground truths' size does not match predictions' size, please check ~ ")
             return 0
         elif len(pre_files) < 1:
             print('No files! Please check~')
@@ -226,19 +226,19 @@ def computeAp(self, label):
             plt.ylabel('recall')
             plt.draw()  # 显示绘图
             # plt.pause(5)  # 显示5秒
-            plt.savefig("class_{}_roc.jpg".format(label))  # 保存图象
+            plt.savefig("class_{}_roc.png".format(label))  # 保存图象
             plt.close()
 
         if self.pr:
-            # 画roc曲线图
+            # 画pr曲线图
             plt.figure('Draw_pr')
             plt.plot(rec, prec)  # plot绘制折线图
             plt.grid(True)
             plt.xlabel('recall')
             plt.ylabel('precision')
             plt.draw()  # 显示绘图
             # plt.pause(5)  # 显示5秒
-            plt.savefig("class_{}_pr.jpg".format(label))  # 保存图象
+            plt.savefig("class_{}_pr.png".format(label))  # 保存图象
             plt.close()
 
         fppi = 0
@@ -276,11 +276,12 @@ def run(self):
         prediction_path, gt_path, predictions, groundtruths, file_format = self.load_all_files()
         aps = 0
 
-        # temp
-        class_map_temp = {1: 'Person', 2: 'Vehicle', 3: 'Dryer'}
+        # modify as you need
+        # list your label names as below ['class 1', 'class 2'......]
+        class_names = ['face']
 
-        for label in range(1, self.cls):
-            semantic_label = class_map_temp[label]
+        for label in class_names:
+            semantic_label = label
             print('Processing label: {}'.format(semantic_label))
             self.get_tp_fp(gt_path, prediction_path, groundtruths, predictions, semantic_label, file_format)
             precision, recall, fppi, fppw, ap = self.computeAp(semantic_label)
@@ -294,8 +295,9 @@ def run(self):
             if self.FPPIW:
                 print('FPPW: ', fppw, 'FPPI', fppi)
             aps += ap
+
         mAp = aps / (self.cls - 1)
-        print ("mAp: ", mAp)
+        print("mAp: ", mAp)
 
         return 0
 

io_file.py

@@ -3,16 +3,24 @@
 
 
 # modify 'class_map' as you need
-class_map = {'Person': 'Person', 'Vehicle': 'Vehicle', 'Dryer': 'Dryer'}
+#           {id or label name in gt label: class name}
+class_map = {'face': 'face'}
 
 
+# parse pascal voc style label file
 def parse_xml(xml_path):
-    dom = xml.dom.minidom.parse(xml_path)
+    gts = []
+    try:
+        dom = xml.dom.minidom.parse(xml_path)
+        print('{} parse failed! Use empty label instead \n'.format(xml_path))
+    except:
+        return gts
     root = dom.documentElement
     objects = root.getElementsByTagName('object')
-    gts = []
     for index, obj in enumerate(objects):
-        name = obj.getElementsByTagName('name')[0].firstChild.data.decode('utf8')
+        name = obj.getElementsByTagName('name')[0].firstChild.data.strip("\ufeff")
+        if name not in class_map:
+            continue
         label = class_map[name]
         bndbox = obj.getElementsByTagName('bndbox')[0]
         x1 = int(bndbox.getElementsByTagName('xmin')[0].firstChild.data)
@@ -21,4 +29,4 @@ def parse_xml(xml_path):
         y2 = int(bndbox.getElementsByTagName('ymax')[0].firstChild.data)
         gt_one = [label, x1, y1, x2, y2]
         gts.append(gt_one)
-    return gts
+    return gts

test.py

@@ -5,14 +5,15 @@
 
 
 cfg = {'file_dir': './',
-       'overlapRatio': 0.5,
-       'cls': 2,
-       'presicion': False,
-       'recall': False,
-       'threshold': 0.5,
-       'FPPIW': False,
-       'roc': False,
-       'pr': False}
+       'overlapRatio': 0.5,  # iou between predicted bounding box and ground truth bounding box
+       'cls': 2,  # background id included
+       'precision': False,  # calculate precision with 'threshold' or not
+       'recall': False,  # calculate precision with 'threshold' or not
+       'threshold': 0.5,  # confidence threshold used in calculating precision and recall
+       'FPPIW': False,  # FPPI: false positive per image; FPPW: false positive per window(bounding box)
+       'roc': False,  # draw roc curve or not
+       'pr': False  # draw pr curve or not
+        }
 
 
 def parse_args():
@@ -43,19 +44,20 @@ def parse_args():
 
 if __name__ == "__main__":
     args = parse_args()
-    args.dir = ['/Users/wangzhe/data/zhengdanongmu/eval/voc_coco_only/converted_voc_coco_only', '/Users/wangzhe/data/zhengdanongmu/eval/voc_coco_only/gt_voc_coco']
+    args.dir = ['/data/guanlang/video_clips/metric_used_xml/face_pred_txt',  # prediction path
+                '/data/guanlang/video_clips/metric_used_xml/gt']  # gt path
 
-    args.cls = 4
-    args.overlapRatio = 0.5
-    args.threshold = 0.5
+    # args.cls = 2
+    # args.overlapRatio = 0.3
+    # args.threshold = 0.94
     len(sys.argv)
-    print ("Your Folder's path: {}".format(args.dir))
-    print ("Overlap Ratio: {}".format(args.overlapRatio))
-    print ("Threshold: {}".format(args.threshold))
-    print ("Num of Categories: {}".format(args.cls))
-    print ("Precision: {}".format(args.precision))
-    print ("Recall: {}".format(args.recall))
-    print ("FPPIW: {}".format(args.FPPIW))
+    print("Your Folder's path: {}".format(args.dir))
+    print("Overlap Ratio: {}".format(args.overlapRatio))
+    print("Threshold: {}".format(args.threshold))
+    print("Num of Categories: {}".format(args.cls))
+    print("Precision: {}".format(args.precision))
+    print("Recall: {}".format(args.recall))
+    print("FPPIW: {}".format(args.FPPIW))
 
     print("Calculating......")