python-wml 3.0.0__py3-none-any.whl

wml/iotoolkit/mpii.py

@@ -0,0 +1,108 @@
+from wml.iotoolkit.mat_data import MatData
+import wml.object_detection2.keypoints as odk
+import numpy as np
+import wml.object_detection2.bboxes as odb
+import math
+'''
+id - joint id (0 - r ankle, 1 - r knee, 2 - r hip, 
+3 - l hip, 4 - l knee, 5 - l ankle, 6 - pelvis, 
+7 - thorax, 8 - upper neck, 9 - head top,
+10 - r wrist, 11 - r elbow, 12 - r shoulder, 13 - l shoulder, 14 - l elbow, 15 - l wrist)
+'''
+def read_mpii_data(path):
+    mpii_data = MatData(path).RELEASE
+    annolist = mpii_data.annolist
+    img_train = mpii_data.img_train
+
+    res = []
+    for i in range(0,len(annolist)):
+        annorect_item = annolist.annorect.get_array_item(i)
+        if annorect_item is None:
+            continue
+        image = str(annolist.image[i].name)
+        image_datas = []
+        try:
+            for j in range(len(annorect_item)):
+                person_anno = annorect_item[j]
+                x1 = person_anno.x1
+                y1 = person_anno.y1
+                x2 = person_anno.x2
+                y2 = person_anno.y2
+                if 'annopoints' not in person_anno.data_keys():
+                    continue
+                points = person_anno.annopoints
+                if points is None:
+                    continue
+                kps = person_anno.annopoints.point
+                res_kps = np.zeros([16,3],dtype=np.float32)
+                for k in range(len(kps)):
+                    kp = kps[k]
+                    id = kp.id
+                    res_kps[id,0] = kp.x
+                    res_kps[id,1] = kp.y
+                    is_visible = kp.is_visible
+                    res_kps[id,2] = float(is_visible)+1 if is_visible is not None else 0
+                person_data = [np.array([x1,y1,x2,y2]),res_kps]
+                image_datas.append(person_data)
+        except Exception as e:
+            if img_train.data[i]==1:
+                print(f"ERROR",i,e,img_train.data[i])
+        if len(image_datas)>0:
+            image_datas = list(zip(*image_datas))
+            bboxes = np.array(image_datas[0])
+            head_bboxes = bboxes
+            kps = np.array(image_datas[1])
+            _bboxes = odk.npbatchget_bboxes(kps)
+            r_bboxes = []
+            for bbox0,bbox1 in zip(bboxes,_bboxes):
+                bbox = odb.bbox_of_boxes([bbox0,bbox1])
+                r_bboxes.append(bbox)
+            bboxes = np.array(r_bboxes,dtype=np.float32)
+            res.append([image,bboxes,kps,head_bboxes])
+
+    return res
+
+class Trans2COCO:
+    def __init__(self) -> None:
+        self.dst_idxs = [5,6,7,8,9,10,11,12,13,14,15,16]
+        self.src_idxs = [13,12,14,11,15,10,3,2,4,1,5,0]
+        self.coco_idxs = [0,1,2,3,4]
+    
+    @staticmethod 
+    def kps_in_bbox(kps,bbox):
+        nr = kps.shape[0]
+        for i in range(nr):
+            if kps[i,2]<= 0:
+                return False
+            x = kps[i,0]
+            y = kps[i,1]
+            if x<bbox[0] or x>bbox[2] or y<bbox[1] or y>bbox[3]:
+                return False
+        return True
+
+    def __call__(self,mpii_kps,coco_kps,head_bboxes):
+        if len(mpii_kps.shape)==2:
+            return self.trans_one(mpii_kps,coco_kps,head_bboxes)
+        res = []
+        for mp,coco,head_bbox in zip(mpii_kps,coco_kps,head_bboxes):
+            res.append(self.trans_one(mp,coco,head_bbox))
+        return np.array(res)
+
+    def trans_one(self,mpii_kps,coco_kps,head_bbox):
+        '''
+        img: [RGB]
+        '''
+        res = np.zeros([17,3],dtype=np.float32)
+        res[self.dst_idxs] = mpii_kps[self.src_idxs]
+        if coco_kps is not None:
+            left_right_pairs = [[5,6],[11,12]]
+            is_good = True
+            for pair in left_right_pairs:
+                l,r = pair
+                if res[l,0]<res[r,0] or res[l,2]<0.1 or res[r,2]<0.1:
+                    is_good = False
+                    break
+
+            if is_good and head_bbox is not None and self.kps_in_bbox(coco_kps[self.coco_idxs],head_bbox):
+                res[self.coco_idxs] = coco_kps[self.coco_idxs]
+        return res

wml/iotoolkit/npmckeypoints_dataset.py

@@ -0,0 +1,164 @@
+#coding=utf-8
+import numpy as np
+import matplotlib.image as mpimg
+import xml.etree.ElementTree as ET
+from xml.dom.minidom import Document
+import random
+import os
+import math
+import wml.wml_utils
+import logging
+import shutil
+from functools import partial
+import wml.wml_utils as wmlu
+import wml.img_utils as wmli
+import copy
+from wml.object_detection2.standard_names import *
+import pickle
+from .common import *
+import wml.object_detection2.bboxes as odb
+from .pascal_voc_toolkit_fwd import *
+from wml.wstructures import WMCKeypoints
+from easydict import EasyDict
+
+
+class NPMCKeypointsDataset(object):
+    def __init__(self, label_text2id=None, shuffle=False,
+                 filter_empty_files=False,
+                 resample_parameters=None,
+                 ignore_case=True):
+        '''
+
+        :param label_text2id: trans a single label text to id:  int func(str)
+        :param shuffle:
+        :param image_sub_dir:
+        :param xml_sub_dir:
+        可以使用datasets_trans/trans_labelme_to_npkp.py将labelme转换为NPMCKeypointsDataset可用的格式
+        '''
+        self.files = None
+        self.shuffle = shuffle
+        self.filter_empty_files = filter_empty_files
+        if isinstance(label_text2id,dict):
+            if ignore_case:
+                new_dict = dict([(k.lower(),v) for k,v in label_text2id.items()])
+                self.label_text2id = partial(ignore_case_dict_label_text2id,dict_data=new_dict)
+            else:
+                self.label_text2id = partial(dict_label_text2id,dict_data=label_text2id)
+        else:
+            self.label_text2id = label_text2id
+
+        if resample_parameters is not None:
+            self.resample_parameters = {}
+            for k,v in resample_parameters.items():
+                if isinstance(k,(str,bytes)):
+                    k = self.label_text2id(k)
+                self.resample_parameters[k] = v
+            print("resample parameters")
+            wmlu.show_dict(self.resample_parameters)
+        else:
+            self.resample_parameters = None
+
+
+    def __len__(self):
+        return len(self.files)
+
+    def __getitem__(self,idx):
+        try:
+            bin_file = self.files[idx]
+        except Exception as e:
+            print(f"ERROR: {e} {self.files[idx]}")
+            print(self.files)
+        #print(xml_file)
+        if not os.path.exists(bin_file):
+            return None
+        try:
+            with open(bin_file,"rb") as f:
+                data = pickle.load(f)
+                labels_names = data[GT_LABELS]
+                img_info = data[IMG_INFO]
+            if self.label_text2id is not None:
+                labels = [self.label_text2id(x) for x in labels_names]
+                keypoints = data[GT_KEYPOINTS]
+                keep = [x is not None for x in labels]
+                labels = [x if x is not None else -1 for x in labels]
+                labels = np.array(labels,dtype=np.int32)
+                labels = labels[keep]
+                keypoints = [keypoints[i] for i in np.where(keep)[0]]
+                data[GT_LABELS] = labels
+                data[GT_KEYPOINTS] = WMCKeypoints(keypoints,width=img_info[WIDTH],height=img_info[HEIGHT])
+            else:
+                labels = None
+
+        except Exception as e:
+            print(f"Read {bin_file} {e} faild.")
+            return None
+        return EasyDict(data)
+
+    def read_data(self,dir_path,suffix=".bin"):
+        if isinstance(dir_path,str):
+            print(f"Read {dir_path}")
+            if not os.path.exists(dir_path):
+                print(f"Data path {dir_path} not exists.")
+                return False
+            self.files = wmlu.get_files(dir_path,suffix=suffix)
+        elif isinstance(dir_path,(list,tuple)) and isinstance(dir_path[0],(str,bytes)) and os.path.isdir(dir_path[0]):
+            self.files = self.get_files_from_dirs(dir_path,
+                                                  suffix=suffix)
+        else:
+            self.files = dir_path
+        if self.filter_empty_files and self.label_text2id:
+            self.files = self.apply_filter_empty_files(self.files)
+        if self.resample_parameters is not None and self.label_text2id:
+            self.files = self.resample(self.files)
+        
+        if len(self.files) == 0:
+            return False
+
+        if self.shuffle:
+            random.shuffle(self.files)
+
+        return True
+
+    def get_files_from_dirs(self,dirs,suffix=".bin"):
+        all_files = []
+        for dir_path in dirs:
+            files = wmlu.get_files(dir_path,
+                                 suffix=suffix)
+            all_files.extend(files)
+
+        return all_files
+    
+    def apply_filter_empty_files(self,files):
+        new_files = []
+        for fs in files:
+            with open(fs,"rb") as f:
+                data = pickle.load(f)
+                labels_name = data[GT_LABELS]
+            labels = [self.label_text2id(x) for x in labels_name]
+            is_none = [x is None for x in labels]
+            if not all(is_none):
+                new_files.append(fs)
+            else:
+                print(f"File {fs} is empty, labels names {labels_name}, labels {labels}")
+
+        return new_files
+    
+    def resample(self,files):
+        all_labels = []
+        for fs in files:
+            with open(fs,"rb") as f:
+                data = pickle.load(f)
+                labels_name = data[GT_LABELS]
+            labels = [self.label_text2id(x) for x in labels_name]
+            all_labels.append(labels)
+
+        return resample(files,all_labels,self.resample_parameters)
+
+
+    def get_items(self):
+        '''
+        :return: 
+        full_path,img_size,category_ids,category_names,boxes,binary_masks,area,is_crowd,num_annotations_skipped
+        '''
+        for i in range(len(self.files)):
+            yield self.__getitem__(i)

wml/iotoolkit/o365_to_coco.py

@@ -0,0 +1,136 @@
+o365_id_to_coco_id = {1:1, # Person:person,  57827
+3:62, # Chair:chair,  9319
+6:3, # Car:car,  4560
+9:44, # Bottle:bottle,  10983
+10:67, # Desk:dining table,  3614
+11:47, # Cup:cup,  12214
+14:31, # Handbag/Satchel:handbag,  1644
+16:51, # Plate:bowl,  4573
+19:84, # Book:book,  2390
+22:9, # Boat:boat,  1939
+25:15, # Bench:bench,  827
+26:64, # Potted Plant:potted plant,  2292
+27:51, # Bowl/Basin:bowl,  12888
+31:86, # Vase:vase,  1162
+35:3, # SUV:car,  1149
+36:46, # Wine Glass:wine glass,  2350
+38:72, # Moniter/TV:tv,  7522
+39:27, # Backpack:backpack,  510
+40:28, # Umbrella:umbrella,  1664
+41:10, # Traffic Light:traffic light,  2109
+44:32, # Tie:tie,  1310
+47:2, # Bicycle:bicycle,  1658
+48:62, # Stool:chair,  2881
+50:3, # Van:car,  968
+51:63, # Couch:couch,  1424
+56:6, # Bus:bus,  730
+57:16, # Wild Bird:bird,  3584
+59:4, # Motorcycle:motorcycle,  1117
+62:77, # Cell Phone:cell phone,  2574
+63:54, # Bread:sandwich,  1396
+65:44, # Canned:bottle,  957
+66:8, # Truck:truck,  1350
+71:88, # Stuffed Toy:teddy bear,  992
+73:9, # Sailboat:boat,  650
+74:73, # Laptop:laptop,  3936
+76:65, # Bed:bed,  1663
+79:19, # Horse:horse,  1071
+82:81, # Sink:sink,  8335
+83:53, # Apple:apple,  4267
+85:49, # Knife:knife,  5252
+88:8, # Pickup Truck:truck,  617
+89:48, # Fork:fork,  5267
+93:18, # Dog:dog,  1075
+94:50, # Spoon:spoon,  4535
+95:85, # Clock:clock,  1289
+97:21, # Cow:cow,  509
+98:61, # Cake:cake,  1834
+99:67, # Dinning Table:dining table,  2437
+100:20, # Sheep:sheep,  460
+105:55, # Orange/Tangerine:orange,  3676
+106:44, # Toiletry:bottle,  6567
+107:76, # Keyboard:keyboard,  5054
+110:8, # Machinery Vehicle:truck,  1010
+113:52, # Banana:banana,  5140
+114:40, # Baseball Glove:baseball glove,  110
+115:5, # Airplane:airplane,  4769
+116:74, # Mouse:mouse,  4689
+117:7, # Train:train,  1270
+119:37, # Soccer:sports ball,  528
+120:35, # Skiboard:skis,  204
+121:33, # Luggage:suitcase,  1321
+127:3, # Sports Car:car,  2115
+128:13, # Stop Sign:stop sign,  435
+129:61, # Dessert:cake,  3077
+130:4, # Scooter:motorcycle,  1518
+133:75, # Remote:remote,  1455
+134:82, # Refrigerator:refrigerator,  2257
+135:79, # Oven:oven,  1765
+136:55, # Lemon:orange,  1638
+137:16, # Duck:bird,  8646
+138:39, # Baseball Bat:baseball bat,  254
+140:17, # Cat:cat,  1374
+142:56, # Broccoli:broccoli,  3389
+144:59, # Pizza:pizza,  1504
+145:22, # Elephant:elephant,  2502
+146:41, # Skateboard:skateboard,  143
+147:42, # Surfboard:surfboard,  1026
+151:60, # Donut:donut,  2980
+152:32, # Bow Tie:tie,  522
+153:57, # Carrot:carrot,  3960
+154:70, # Toilet:toilet,  3117
+155:38, # Kite:kite,  1379
+157:37, # Other Balls:sports ball,  1149
+162:44, # Cleaning Products:bottle,  2660
+164:78, # Microwave:microwave,  2085
+165:16, # Pigeon:bird,  3948
+166:37, # Baseball:sports ball,  509
+168:67, # Coffee Table:dining table,  44
+170:87, # Scissors:scissors,  1488
+172:54, # Pie:sandwich,  516
+174:36, # Snowboard:snowboard,  274
+177:11, # Fire Hydrant:fire hydrant,  222
+178:37, # Basketball:sports ball,  363
+179:24, # Zebra:zebra,  2019
+181:25, # Giraffe:giraffe,  1084
+189:8, # Fire Truck:truck,  525
+190:37, # Billards:sports ball,  2809
+193:62, # Wheelchair:chair,  40*
+195:33, # Briefcase:suitcase,  764
+200:8, # Heavy Truck:truck,  639
+201:54, # Hamburger:sandwich,  1227
+205:43, # Tennis Racket:tennis racket,  193
+207:37, # American Football:sports ball,  165
+208:16, # earphone:bird,  9
+211:37, # Tennis:sports ball,  523
+212:9, # Ship:boat,  1004
+220:34, # Frisbee:frisbee,  330
+222:16, # Chicken:bird,  2043
+227:90, # Toothbrush:toothbrush,  2152
+235:58, # Hot dog:hot dog,  547
+237:53, # Peach:apple,  1085
+240:37, # Volleyball:sports ball,  198
+242:16, # Goose:bird,  2717
+248:37, # Golf Ball:sports ball,  565
+249:8, # Ambulance:truck,  715
+250:14, # Parking meter:parking meter,  59
+258:16, # Penguin:bird,  1751
+263:16, # Swan:bird,  1863
+264:5, # Helicopter:airplane,  114
+266:54, # Sandwich:sandwich,  789
+278:80, # Toaster:toaster,  498
+281:61, # Cheese:cake,  200
+291:44, # Flask:bottle,  84
+296:23, # Bear:bear,  133
+309:3, # Formula 1 :car,  199
+310:53, # Pomegranate:apple,  266
+319:20, # Antelope:sheep,  62
+320:16, # Parrot:bird,  342
+
+326:70, # Urinal:toilet,  460
+329:89, # Hair Dryer:hair drier,  315
+330:61, # Egg tart:cake,  150
+345:21, # Yak:cow,  413
+356:90, # Cosmetics Brush/Eyeliner Pencil:toothbrush,  5
+365:37, # Table Tennis :sports ball,  13
+}

wml/iotoolkit/object365_toolkit.py

@@ -0,0 +1,156 @@
+import json
+import os.path as osp
+import wml.object_detection2.bboxes as odb
+import numpy as np
+import wml.object_detection2.visualization as odv
+import wml.wml_utils as wmlu
+import PIL.Image as Image
+import wml.img_utils as wmli
+from .common import *
+
+
+class Object365:
+    def __init__(self,absolute_coord=True):
+        self.json_path = None
+        self.img_dir = None
+        self.data = None
+        self.absolute_coord = absolute_coord
+        self.images_data = {}
+        self.annotation_data = {}
+        self.ids = []
+        self.id2name = {}
+
+    def read_data(self,json_path,img_dir):
+        self.json_path = json_path
+        self.img_dir = img_dir
+        with open(self.json_path,"r") as f:
+            self.data = json.load(f)
+        self.images_data = {}
+        self.ids = []
+        self.annotation_data = {}
+
+        for d in self.data['images']:
+            id = d['id']
+            self.ids.append(id)
+            self.images_data[id] = d
+        for d in self.data['annotations']:
+            id = d['image_id']
+            self.add_anno_data(self.annotation_data,id,d)
+        name_dict = {}
+        for x in self.data['categories']:
+            name_dict[x['id']] = x['name']
+        self.id2name = name_dict
+        info = list(name_dict.items())
+        info.sort(key=lambda x:x[0])
+        print(f"Category")
+        wmlu.show_list(info)
+
+    @staticmethod
+    def add_anno_data(dict_data,id,item_data):
+        if id in dict_data:
+            dict_data[id].append(item_data)
+        else:
+            dict_data[id] = [item_data]
+
+    def __len__(self):
+        return len(self.ids)
+
+    def getitem_by_id(self,id):
+        idx = self.ids.index(id)
+        return self.__getitem__(idx)
+
+    def __getitem__(self, item):
+        try:
+            id = self.ids[item]
+            item_data = self.annotation_data[id]
+            is_crowd = []
+            bboxes = []
+            labels = []
+            labels_names = []
+            for data in item_data:
+                is_crowd.append(data['iscrowd'])
+                bboxes.append(data['bbox'])
+                label = data['category_id']
+                label_text = self.id2name[label]
+                labels.append(label)
+                labels_names.append(label_text)
+            bboxes = np.array(bboxes,dtype=np.float32)
+            bboxes[...,2:] = bboxes[...,:2]+bboxes[...,2:]
+            bboxes = odb.npchangexyorder(bboxes)
+            is_crowd = np.array(is_crowd,dtype=np.float32)
+            image_data = self.images_data[id]
+            shape = [image_data['height'],image_data['width']]
+            if not self.absolute_coord:
+                bboxes = odb.absolutely_boxes_to_relative_boxes(bboxes,width=shape[1],height=shape[0])
+            img_name = image_data['file_name']
+            img_file = osp.join(self.img_dir,img_name)
+            '''
+            bboxes: [N,4] (y0,x0,y1,x1)
+            '''
+            return DetData(img_file, shape, labels, labels_names, bboxes, None, None, is_crowd, None)
+        except Exception as e:
+            print(e)
+            return None
+
+    def get_items(self):
+        for i in range(len(self.ids)):
+            res = self.__getitem__(i)
+            if res is None:
+                continue
+            yield res
+
+class TorchObject365(Object365):
+    def __init__(self, img_dir,anno_path,absolute_coord=True):
+        super().__init__(absolute_coord)
+        super().read_data(anno_path,img_dir)
+
+    def __getitem__(self, item):
+        x = super().__getitem__(item)
+        full_path, shape,category_ids, category_names, boxes, binary_mask, area, is_crowd, num_annotations_skipped = x
+        img = wmli.imread(full_path)
+        img = Image.fromarray(img)
+        res = []
+        nr = len(category_ids)
+        boxes = odb.npchangexyorder(boxes)
+        boxes[...,2:] = boxes[...,2:] - boxes[...,:2]
+        for i in range(nr):
+            item = {"bbox":boxes[i],"category_id":category_ids[i],"iscrowd":is_crowd[i]}
+            res.append(item)
+        
+        return img,res
+
+if __name__ == "__main__":
+    import wml.img_utils as wmli
+    import random
+    import matplotlib.pyplot as plt
+    import time
+
+    random.seed(time.time())
+
+    save_dir = "/home/wj/ai/mldata1/Objects365/tmp"
+
+    wmlu.create_empty_dir_remove_if(save_dir)
+    data = Object365(absolute_coord=False)
+    data.read_data("/home/wj/ai/mldata1/Objects365/Annotations/train/train.json","/home/wj/ai/mldata1/Objects365/Images/train/train")
+    idxs = list(range(len(data)))
+    random.shuffle(idxs)
+    max_nr = 100
+    idxs = idxs[:max_nr]
+    for idx in idxs:
+        x = data[idx]
+        full_path, shape,category_ids, category_names, boxes, binary_mask, area, is_crowd, num_annotations_skipped = x
+        img = wmli.imread(full_path)
+
+        def text_fn(classes, scores):
+            return data.id2name[classes]
+
+        img = odv.draw_bboxes(
+            img=img, classes=category_ids, scores=None, bboxes=boxes, color_fn=None,
+            text_fn=text_fn, thickness=2,
+            show_text=True,
+            font_scale=0.8)
+        save_path = osp.join(save_dir,osp.basename(full_path))
+        wmli.imwrite(save_path,img)
+        '''plt.figure()
+        plt.imshow(img)
+        plt.show()'''

wml/iotoolkit/object365v2_toolkit.py

@@ -0,0 +1,71 @@
+from wml.iotoolkit.coco_toolkit import *
+import os.path as osp
+from .o365_to_coco import *
+from .coco_data_fwd import ID_TO_TEXT
+from functools import partial
+
+
+def trans_file_name(filename,image_dir):
+    '''
+    object365文件名使用了类似于:'images/v1/patch8/objects365_v1_00420917.jpg'格式，而实际存的为
+    data_root/train/pathch8/objects365_v1_00420917.jpg这种格式，这里仅保留最后的目录名及文件名，也就是输出
+    pathch8/objects365_v1_00420917.jpg
+    注: read_data的image_dir需要设置为data_root/train/
+    '''
+    names = filename.split("/")[-2:]
+    return osp.join(*names)
+
+def trans_label2coco(id,fn=None):
+    if id in o365_id_to_coco_id:
+        id = o365_id_to_coco_id[id]
+        if fn is not None:
+            return fn(id)
+        else:
+            return id
+    else:
+        return None
+
+class Object365V2(COCOData):
+    def __init__(self,is_relative_coordinate=False,trans2coco=False,remove_crowd=True,trans_label=None):
+        super().__init__(is_relative_coordinate=is_relative_coordinate,remove_crowd=remove_crowd,trans_label=trans_label,include_masks=False)
+        if trans2coco and trans_label is not None:
+            self.trans_label = partial(trans_label2coco,fn=trans_label)
+        elif trans2coco:
+            self.trans_label = trans_label2coco
+        elif trans_label is not None:
+            self.trans_label = trans_label
+        self.trans_file_name = trans_file_name
+        if trans2coco:
+            self.id2name = {}
+            for k,info in ID_TO_TEXT.items():
+                self.id2name[k] = info['name']
+
+    def read_data(self,annotations_file,image_dir=None):
+        if image_dir is None:
+            image_dir = osp.dirname(osp.abspath(annotations_file))
+        return super().read_data(annotations_file,image_dir)
+
+class TorchObject365V2(Object365V2):
+    def __init__(self, img_dir, anno_path,trans2coco=False,trans_label=None):
+        super().__init__(is_relative_coordinate=False,trans2coco=trans2coco,trans_label=trans_label)
+        super().read_data(anno_path, img_dir)
+
+    def __getitem__(self, item):
+        x = super().__getitem__(item)
+        full_path, shape, category_ids, category_names, boxes, binary_mask, area, is_crowd, num_annotations_skipped = x
+        try:
+            img = wmli.imread(full_path)
+        except Exception as e:
+            print(f"Read {full_path} faild, error:{e}")
+            img = np.zeros([shape[0],shape[1],3],dtype=np.uint8)
+        img = Image.fromarray(img)
+        res = []
+        nr = len(category_ids)
+        boxes = odb.npchangexyorder(boxes)
+        boxes[..., 2:] = boxes[..., 2:] - boxes[..., :2]
+        #new bboxes is [N,4], [x0,y0,w,h]
+        for i in range(nr):
+            item = {"bbox": boxes[i], "category_id": category_ids[i], "iscrowd": is_crowd[i],"area":area[i]}
+            res.append(item)
+
+        return img, res

wml/iotoolkit/pascal_voc_data.py

@@ -0,0 +1,51 @@
+#coding=utf-8
+ID_TO_TEXT= {
+  1:'aeroplane',
+  2:'bicycle',
+  3:'bird',
+  4:'boat',
+  5:'bottle',
+  6:'bus',
+  7:'car',
+  8:'cat',
+  9:'chair',
+  10:'cow',
+  11:'diningtable',
+  12:'dog',
+  13:'horse',
+  14:'motorbike',
+  15:'person',
+  16:'pottedplant',
+  17:'sheep',
+  18:'sofa',
+  19:'train',
+  20:'tvmonitor'
+}
+TEXT_TO_ID = {
+    'aeroplane':  1,
+    'bicycle':  2,
+    'bird':  3,
+    'boat':  4,
+    'bottle':  5,
+    'bus':  6,
+    'car':  7,
+    'cat':  8,
+    'chair':  9,
+    'cow':  10,
+    'diningtable':  11,
+    'dog':  12,
+    'horse':  13,
+    'motorbike':  14,
+    'person':  15,
+    'pottedplant':  16,
+    'sheep':  17,
+    'sofa':  18,
+    'train':  19,
+    'tvmonitor':20
+}
+def get_id_to_textv2():
+    res = {}
+    for k,v in ID_TO_TEXT.items():
+        res[k] = {"name":v}
+    return res
+