add csv2labelme

Author: zhuchaojie

README.md

@@ -3,6 +3,14 @@
 万事开头难。之前写图像识别的博客教程，也是为了方便那些学了很多理论知识，却对实际项目无从下手的小伙伴，后来转到目标检测来了，师从烨兄、亚光兄，从他们那学了不少检测的知识和操作，今天也终于闲下了，准备写个检测系列的总结。一方面分享知识希望可以一起学习，另一方面让一部分人少走弯路，快速上路（入坑）。
 
 此部分代码：[Github](https://github.com/spytensor/prepare_detection_dataset)
+博客地址: [目标检测系列一：如何制作数据集?](http://www.spytensor.com/index.php/archives/48/)
+
+
+**更新**
+
+- (28/03/2019)
+    - 新增 `csv2labelme`
+
 
 <h4 id="1">1. 内容介绍</h4>
 
@@ -214,6 +222,18 @@ saved_coco_path = "./"     # path for saved coco dataset
 ```
 然后运行 `python labelme2voc.py`，生成文件形式同`csv2voc`
 
+<h5 id="3.5">3.5 csv2labelme</h5>
+
+首先更改`csv2labelme.py`中以下几个配置
+
+```
+image_path = "./images/"  # path for images
+csv_file = "./"     # path for csv annotations
+```
+然后运行 `python csv2labelme.py`，生成的`json`文件会保存在`image_path`下,切换路径过去,直接`labelme`便
+可以查看标签.
+
+
 <h4 id="4">4. 万能中介csv</h4>
 
 从上面的转换格式中可以看出，并没有给出如何转到csv的，一是因为太过于简单，而是主流检测框架很少支持这种格式的数据输入。以下给出如何将标注信息写入`csv`

csv2labelme.py

@@ -0,0 +1,49 @@
+import os
+import cv2
+import json
+import pandas as pd
+import numpy as np
+from glob import glob 
+from tqdm import tqdm
+from IPython import embed
+import base64
+from labelme import utils
+image_path = "./images/"
+csv_file = "./train_labels.csv"
+annotations = pd.read_csv(csv_file,header=None).values
+total_csv_annotations = {}
+for annotation in annotations:
+    key = annotation[0].split(os.sep)[-1]
+    value = np.array([annotation[1:]])
+    if key in total_csv_annotations.keys():
+        total_csv_annotations[key] = np.concatenate((total_csv_annotations[key],value),axis=0)
+    else:
+        total_csv_annotations[key] = value
+for key,value in total_csv_annotations.items():
+    height,width = cv2.imread(image_path+key).shape[:2]
+    labelme_format = {
+    "version":"3.6.16",
+    "flags":{},
+    "lineColor":[0,255,0,128],
+    "fillColor":[255,0,0,128],
+    "imagePath":key,
+    "imageHeight":height,
+    "imageWidth":width
+    }
+    with open(image_path+key,"rb") as f:
+        imageData = f.read()
+        imageData = base64.b64encode(imageData).decode('utf-8')
+    #img = utils.img_b64_to_arr(imageData)
+    labelme_format["imageData"] = imageData
+    shapes = []
+    for shape in value:
+        label = shape[-1]
+        s = {"label":label,"line_color":None,"fill_color":None,"shape_type":"rectangle"}
+        points = [
+            [shape[0],shape[1]],
+            [shape[2],shape[3]]
+            ]
+        s["points"] = points
+        shapes.append(s)
+    labelme_format["shapes"] = shapes
+    json.dump(labelme_format,open("%s/%s/"%(image_path,key.replace(".jpg",".json")),"w"),ensure_ascii=False, indent=2)