Fruits 360 dataset

Fruits 360 dataset: A dataset of images containing fruits Version: 2018.09.07.0 The following fruits are included: Apples (different varieties: Golden, Golden-Red, Granny Smith, Red, Red Delicious), Apricot, Avocado, Avocado ripe, Banana (Yellow, Red), Cactus fruit, Cantaloupe (2 varieties), Carambula, Cherry (different varieties, Rainier), Cherry Wax (Yellow, Red, Black), Clementine, Cocos, Dates, Granadilla, Grape (Pink, White, White2), Grapefruit (Pink, White), Guava, Huckleberry, Kiwi, Kaki, Kumsquats, Lemon (normal, Meyer), Lime, Lychee, Mandarine, Mango, Maracuja, Melon Piel de Sapo, Mulberry, Nectarine, Orange, Papaya, Passion fruit, Peach, Pepino, Pear (different varieties, Abate, Monster, Williams), Physalis (normal, with Husk), Pineapple (normal, Mini), Pitahaya Red, Plum, Pomegranate, Quince, Rambutan, Raspberry, Salak, Strawberry (normal, Wedge), Tamarillo, Tangelo, Tomato (different varieties, Maroon, Cherry Red), Walnut. Dataset properties Total number of images: 55244. Training set size: 41322 images (one fruit per image). Test set size: 13877 images (one fruit per image). Multi-fruits set size: 45 images (more than one fruit (or fruit class) per image) Number of classes: 81 (fruits). Image size: 100x100 pixels. Filename format: image_index_100.jpg (e.g. 32_100.jpg) or r_image_index_100.jpg (e.g. r_32_100.jpg) or r2_image_index_100.jpg. "r" stands for rotated fruit. "r2" means that the fruit was rotated around the 3rd axis. "100" comes from image size (100x100 pixels). Different varieties of the same fruit (apple for instance) are stored as belonging to different classes. How we made it Fruits were planted in the shaft of a low speed motor (3 rpm) and a short movie of 20 seconds was recorded. A Logitech C920 camera was used for filming the fruits. This is one of the best webcams available. Behind the fruits we placed a white sheet of paper as background. However due to the variations in the lighting conditions, the background was not uniform and we wrote a dedicated algorithm which extract the fruit from the background. This algorithm is of flood fill type: we start from each edge of the image and we mark all pixels there, then we mark all pixels found in the neighborhood of the already marked pixels for which the distance between colors is less than a prescribed value. We repeat the previous step until no more pixels can be marked. All marked pixels are considered as being background (which is then filled with white) and the rest of pixels are considered as belonging to the object. The maximum value for the distance between 2 neighbor pixels is a parameter of the algorithm and is set (by trial and error) for each movie. Published research papers Horea Muresan, Mihai Oltean, Fruit recognition from images using deep learning, Acta Univ. Sapientiae, Informatica Vol. 10, Issue 1, pp. 26-42, 2018. License MIT License Copyright (c) 2017-2018 Mihai Oltean, Horea Muresan

Fruits 360 数据集：一款涵盖各类水果图像的数据集，版本为2018.09.07.0。
本次数据集涵盖的水果品类如下：苹果（含多个品种：金冠、金红、澳洲青苹（Granny Smith）、红苹果、蛇果（Red Delicious））、杏、牛油果、成熟牛油果、香蕉（黄色、红色）、仙人掌果、哈密瓜（2个品种）、杨桃（Carambula）、樱桃（含多个品种，雷尼尔樱桃（Rainier））、蜡质樱桃（黄色、红色、黑色）、克莱门氏小柑橘、椰子、椰枣、龙珠果（Granadilla）、葡萄（粉色、白色、白色2）、西柚（粉色、白色）、番石榴、越橘（Huckleberry）、猕猴桃、柿子、金桔（Kumsquats）、柠檬（普通品种、迈耶柠檬（Meyer））、青柠、荔枝、柑橘、芒果、西番莲果（Maracuja）、萨波甜瓜（Melon Piel de Sapo）、桑葚、油桃、橙子、番木瓜、百香果（Passion fruit）、桃、香瓜茄（Pepino）、梨（含多个品种：阿巴特、莫纳斯特、威廉姆斯）、酸浆（Physalis，普通品种、带壳品种）、菠萝（普通品种、迷你品种）、红心火龙果（Pitahaya Red）、李子、石榴、榅桲（Quince）、红毛丹、树莓、蛇皮果（Salak）、草莓（普通品种、楔形切块品种）、树番茄（Tamarillo）、坦格罗橘柚（Tangelo）、番茄（含多个品种：酱红色、樱桃红）、胡桃。
数据集属性：
总图像量：55244张。
训练集规模：41322张图像（单张图像仅包含一个水果）。
测试集规模：13877张图像（单张图像仅包含一个水果）。
多果样本集规模：45张图像（单张图像包含多个水果或水果品类）。
类别总数：81类（水果品类）。
图像分辨率：100×100像素。
文件名格式：image_index_100.jpg（例如32_100.jpg），或r_image_index_100.jpg（例如r_32_100.jpg），或r2_image_index_100.jpg。其中“r”代表水果已被旋转，“r2”代表水果沿第三轴旋转；“100”取自图像分辨率100×100像素。
同一水果的不同品种（以苹果为例）会被划分为不同的类别进行存储。
数据集采集流程：
将水果放置于转速为3转每分钟的低速电机转轴上，录制一段20秒的短视频。使用罗技C920网络摄像头（Logitech C920）进行拍摄，该机型为当前主流优质网络摄像头之一。拍摄背景采用白色卡纸，但受光照条件差异影响，背景并非完全均匀，因此我们开发了专用算法实现前景水果与背景的分离。该算法采用泛洪填充（flood fill）逻辑：从图像的每条边缘开始标记像素，随后对已标记像素邻域内、色彩差值低于设定阈值的未标记像素进行标记，重复该过程直至无可标记新像素。所有被标记的像素将被判定为背景并填充为白色，剩余像素则判定为目标物体所属区域。相邻像素间的最大色彩差值为算法参数，需针对每段视频通过试错法进行调整。
相关发表研究论文：
Horea Muresan、Mihai Oltean，《基于深度学习的图像水果识别》，发表于《Acta Univ. Sapientiae, Informatica》第10卷第1期，第26-42页，2018年。
许可证：MIT许可证（MIT License），版权所有© 2017-2018 Mihai Oltean、Horea Muresan。