list readme

README.md

@@ -1,14 +1,73 @@
-# openbbox
+# YOLO v2 Bounding Box Tool
 
-Feel free to contribute!
+[![New](https://img.shields.io/badge/2018-NEW-brightgreen.svg)](https://github.com/Cartucho/yolo-boundingbox-labeler-GUI/commits/master)
+[![GitHub stars](https://img.shields.io/github/stars/Cartucho/yolo-boundingbox-labeler-GUI.svg?style=social&label=Stars)](https://github.com/Cartucho/yolo-boundingbox-labeler-GUI)
 
-An open source bounding box tool using OpenCV 3 and Python.
+Bounding box labeler tool to generate the training data in the format YOLO v2 requires.
 
-Easily generate the training data in the format YOLOv2 requires.
+<img src="https://media.giphy.com/media/l49JDgDSygJN369vW/giphy.gif" width="40%"><img src="https://media.giphy.com/media/3ohc1csRs9PoDgCeuk/giphy.gif" width="40%">
+<img src="https://media.giphy.com/media/3o752fXKwTJJkhXP32/giphy.gif" width="40%"><img src="https://media.giphy.com/media/3ohc11t9auzSo6fwLS/giphy.gif" width="40%">
 
-## Instructions:
+## Table of contents
 
-1. Insert images images in the folder 1.insert_images_here.
-1.1 YOLO only accepts .jpg and .jpeg images, other types will be ignored
-2. Insert class list in the file 2.insert_class_list_here.txt
-3. Run the code: python run_this.py
+- [Quick start](#quick-start)
+- [Prerequisites](#prerequisites)
+- [Run project](#run-project)
+- [GUI usage](#gui-usage)
+- [Authors](#authors)
+
+## Quick start
+
+To start using the YOLO Bounding Box Tool you need to [download the latest release](https://github.com/Cartucho/yolo-boundingbox-labeler-GUI/archive/v1.1.zip) or clone the repo:
+
+```
+git clone https://github.com/Cartucho/yolo-boundingbox-labeler-GUI
+```
+
+### Prerequisites
+
+You need to install:
+
+- [OpenCV](https://opencv.org/) version >= 3.0
+  - [Installation in Windows](https://docs.opencv.org/master/d3/d52/tutorial_windows_install.html)
+  - [Installation in Linux](https://docs.opencv.org/master/d7/d9f/tutorial_linux_install.html)
+- [Python](https://www.python.org/downloads/)
+
+### Run project
+
+Step by step:
+
+  1. Insert the images in the folder **images/**
+  2. Insert the class list in the file **class_list.txt**
+  3. Run the code:
+         ```
+         python run.py
+         ```
+  4. You can find the bounding box files in the folder **bbox_txt/**
+
+### GUI usage
+
+Keyboard, press: 
+
+<img src="https://github.com/Cartucho/yolo-boundingbox-labeler-GUI/blob/master/keyboard_usage.jpg">
+
+| Key | Description |
+| --- | --- |
+| h | help |
+| q | quit |
+| e | edges |
+| a/d | previous/next image |
+| s/w | previous/next class |
+
+
+Mouse:
+  - Use two left clicks to do each bounding box
+  - Use the middle mouse to zoom in and out
+
+## Authors
+
+* **João Cartucho** - Please give me your feedback: [email protected]
+
+    Feel free to contribute
+
+    [![GitHub contributors](https://img.shields.io/github/contributors/Cartucho/yolo-boundingbox-labeler-GUI.svg)](https://github.com/Cartucho/yolo-boundingbox-labeler-GUI/graphs/contributors)

3.run_this.py → run.py

@@ -7,9 +7,11 @@
 class_index = 0
 img_index = 0
 img = None
+img_objects = []
+bb_dir = "bbox_txt/"
+
 mouse_x = 0
 mouse_y = 0
-
 point_1 = (-1, -1)
 point_2 = (-1, -1)
 
@@ -72,15 +74,27 @@ def yolo_format(class_index, point_1, point_2, height, width):
 
 
 def get_txt_path(img_path):
+    img_name = img_path.split('/')[-1]
     img_type = img_path.split('.')[-1]
-    return img_path.replace(img_type, 'txt')
+    return bb_dir + img_name.replace(img_type, 'txt')
 
 
-def save_bb(text_path, line):
-    with open(text_path, 'a') as myfile:
+def save_bb(txt_path, line):
+    with open(txt_path, 'a') as myfile:
         myfile.write(line + "\n") # append line
 
 
+def delete_bb(txt_path, line_index):
+    with open(txt_path, "r") as old_file:
+        lines = old_file.readlines()
+
+    with open(txt_path, "w") as new_file:
+        counter = 0
+        for line in lines:
+            if counter is not line_index:
+                new_file.write(line)
+            counter += 1
+
 def yolo_to_x_y(x_center, y_center, x_width, y_height, width, height):
     x_center *= width
     y_center *= height
@@ -91,16 +105,22 @@ def yolo_to_x_y(x_center, y_center, x_width, y_height, width, height):
     return int(x_center - x_width), int(y_center - y_height), int(x_center + x_width), int(y_center + y_height)
 
 def draw_bboxes_from_file(tmp_img, txt_path, width, height):
+    global img_objects
+    img_objects = []
     if os.path.isfile(txt_path):
         with open(txt_path) as f:
             content = f.readlines()
         for line in content:
             values_str = line.split()
             class_index, x_center, y_center, x_width, y_height = map(float, values_str)
+            class_index = int(class_index)
             # convert yolo to points
             x1, y1, x2, y2 = yolo_to_x_y(x_center, y_center, x_width, y_height, width, height)
-            color = class_rgb[int(class_index)]
+            img_objects.append([class_index, x1, y1, x2, y2])
+            color = class_rgb[class_index]
             cv2.rectangle(tmp_img, (x1, y1), (x2, y2), color, 2)
+            font = cv2.FONT_HERSHEY_SIMPLEX
+            cv2.putText(tmp_img, class_list[class_index], (x1, y1 - 5), font, 0.6, color, 2, cv2.LINE_AA)
     return tmp_img
 
 # mouse callback function
@@ -111,22 +131,53 @@ def draw_roi(event, x, y, flags, param):
         mouse_y = y
     elif event == cv2.EVENT_LBUTTONDOWN:
         if point_1[0] is -1:
-            # first click
-            cv2.displayOverlay(WINDOW_NAME, "Currently selected label:\n"
-                                    "" + class_list[class_index] + "", 2000)
+            # first click (start drawing a bounding box or delete an item)
             point_1 = (x, y)
         else:
             # second click
             point_2 = (x, y)
 
+
+def is_mouse_inside_points(x1, y1, x2, y2):
+    return mouse_x > x1 and mouse_x < x2 and mouse_y > y1 and mouse_y < y2
+
+
+def get_close_icon(x1, y1, x2, y2):
+    percentage = 0.1
+    height = -1
+    while height < 15 and percentage < 1.0:
+        height = int((y2 - y1) * percentage)
+        percentage += 0.1
+    return (x2 - height), y1, x2, (y1 + height)
+
+
+def draw_close_icon(tmp_img, x1_c, y1_c, x2_c, y2_c):
+    red = (0,0,255)
+    cv2.rectangle(tmp_img, (x1_c + 1, y1_c - 1), (x2_c, y2_c), red, -1)
+    white = (255, 255, 255)
+    cv2.line(tmp_img, (x1_c, y1_c), (x2_c, y2_c), white, 2)
+    cv2.line(tmp_img, (x1_c, y2_c), (x2_c, y1_c), white, 2)
+    return tmp_img
+
+
+def draw_info_if_bb_selected(tmp_img):
+    for obj in img_objects:
+        ind, x1, y1, x2, y2 = obj
+        if is_mouse_inside_points(x1, y1, x2, y2):
+            x1_c, y1_c, x2_c, y2_c = get_close_icon(x1, y1, x2, y2)
+            tmp_img = draw_close_icon(tmp_img, x1_c, y1_c, x2_c, y2_c)
+    return tmp_img
+
+
 # load img list
-image_list = glob.glob('1.insert_images_here/*.jpg')
-image_list.extend(glob.glob('1.insert_images_here/*.jpeg'))
-print(image_list)
+img_dir = "images/"
+image_list = glob.glob(img_dir +'*.jpg')
+image_list.extend(glob.glob(img_dir + '*.jpeg'))
+#print(image_list)
 last_img_index = len(image_list) - 1
 
 # load class list
-with open('2.insert_class_list_here.txt') as f:
+with open('class_list.txt') as f:
     class_list = f.read().splitlines()
 #print(class_list)
 last_class_index = len(class_list) - 1
@@ -135,7 +186,7 @@ def draw_roi(event, x, y, flags, param):
 class_rgb = np.random.random_integers(low=0, high=255, size=(3,len(class_list)))
 
 # create window
-WINDOW_NAME = 'openbbox'
+WINDOW_NAME = 'Bounding Box Labeler'
 cv2.namedWindow(WINDOW_NAME, cv2.WINDOW_KEEPRATIO)
 cv2.resizeWindow(WINDOW_NAME, 1000, 700)
 cv2.setMouseCallback(WINDOW_NAME, draw_roi)
@@ -152,8 +203,11 @@ def draw_roi(event, x, y, flags, param):
 change_img_index(0)
 edges_on = False
 
-cv2.displayOverlay(WINDOW_NAME, "Welcome to openbbox!\n Press [h] for help.", 4000)
-print(" Welcome to openbbox!\n Select the window and press [h] for help.")
+cv2.displayOverlay(WINDOW_NAME, "Welcome!\n Press [h] for help.", 4000)
+print(" Welcome!\n Select the window and press [h] for help.")
+
+if not os.path.exists(bb_dir):
+    os.makedirs(bb_dir)
 
 # loop
 while True:
@@ -163,28 +217,46 @@ def draw_roi(event, x, y, flags, param):
     if edges_on == True:
         # draw edges
         tmp_img = draw_edges(tmp_img)
-    # draw vertical yellow lines
-    draw_line(tmp_img, mouse_x, mouse_y, width, height) # show vertical and horizontal line
+    # draw vertical and horizong yellow guide lines
+    draw_line(tmp_img, mouse_x, mouse_y, width, height)
     img_path = image_list[img_index]
     txt_path = get_txt_path(img_path)
     # draw already done bounding boxes
     tmp_img = draw_bboxes_from_file(tmp_img, txt_path, width, height)
+    # if bounding box is selected add extra info
+    tmp_img = draw_info_if_bb_selected(tmp_img)
     # if first click
     if point_1[0] is not -1:
-        color = class_rgb[class_index]
-        # draw partial bbox
-        cv2.rectangle(tmp_img, point_1, (mouse_x, mouse_y), color, 2)
-        # if second click
-        if point_2[0] is not -1:
-            # save the bounding box
-            line = yolo_format(class_index, point_1, point_2, width, height)
-            save_bb(txt_path, line)
-            # reset the points
-            point_1 = (-1, -1)
-            point_2 = (-1, -1)
+        removed_an_object = False
+        # if clicked inside a delete button, then remove that object
+        for obj in img_objects:
+            ind, x1, y1, x2, y2 = obj
+            x1, y1, x2, y2 = get_close_icon(x1, y1, x2, y2)
+            if is_mouse_inside_points(x1, y1, x2, y2):
+                # remove that object
+                delete_bb(txt_path, img_objects.index(obj))
+                removed_an_object = True
+                point_1 = (-1, -1)
+                break
+
+        if not removed_an_object:
+            color = class_rgb[class_index]
+            # draw partial bbox
+            cv2.rectangle(tmp_img, point_1, (mouse_x, mouse_y), color, 2)
+            # if second click
+            if point_2[0] is not -1:
+                # save the bounding box
+                line = yolo_format(class_index, point_1, point_2, width, height)
+                save_bb(txt_path, line)
+                # reset the points
+                point_1 = (-1, -1)
+                point_2 = (-1, -1)
+            else:
+                cv2.displayOverlay(WINDOW_NAME, "Selected label: " + class_list[class_index] + ""
+                                        "\nPress [w] or [s] to change.", 120)
 
     cv2.imshow(WINDOW_NAME, tmp_img)
-    pressed_key = cv2.waitKey(50)
+    pressed_key = cv2.waitKey(100)
 
     """ Key Listeners START """
     if pressed_key == ord('a') or pressed_key == ord('d'):