From 62440b5bb91d063be438b67c93eed1aa1f74ce57 Mon Sep 17 00:00:00 2001 From: "li.zhenye" Date: Thu, 21 Jul 2022 16:10:26 +0800 Subject: [PATCH] =?UTF-8?q?=E6=B7=BB=E5=8A=A0=E6=96=87=E6=A1=A3=E5=92=8C?= =?UTF-8?q?=E6=95=B4=E7=90=86?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- 01_dataset.ipynb | 18 +-- 02_classification.ipynb | 235 ++++++++++++++++++++++++---------------- README.md | 66 ++++++++++- 3 files changed, 210 insertions(+), 109 deletions(-) diff --git a/01_dataset.ipynb b/01_dataset.ipynb index f148618..3edb2f2 100644 --- a/01_dataset.ipynb +++ b/01_dataset.ipynb @@ -13,7 +13,7 @@ }, { "cell_type": "code", - "execution_count": 25, + "execution_count": 1, "metadata": { "pycharm": { "name": "#%%\n" @@ -40,7 +40,7 @@ }, { "cell_type": "code", - "execution_count": 26, + "execution_count": 2, "metadata": { "pycharm": { "name": "#%%\n" @@ -49,19 +49,7 @@ "outputs": [], "source": [ "img_path = r\"data/dataset_old/img/yangeng.bmp\"\n", - "label_path = r\"data/dataset_old/label/yangeng.bmp\"" - ] - }, - { - "cell_type": "code", - "execution_count": 27, - "metadata": { - "pycharm": { - "name": "#%%\n" - } - }, - "outputs": [], - "source": [ + "label_path = r\"data/dataset_old/label/yangeng.bmp\"\n", "# 读取图片和色彩空间转换\n", "img = cv2.imread(img_path)\n", "label_img = cv2.imread(label_path)\n", diff --git a/02_classification.ipynb b/02_classification.ipynb index a820e2e..9055ffc 100644 --- a/02_classification.ipynb +++ b/02_classification.ipynb @@ -7,13 +7,11 @@ "name": "#%% md\n" } }, - "source": [ - "# 模型的训练" - ] + "source": [] }, { "cell_type": "code", - "execution_count": 1, + "execution_count": 5, "metadata": { "pycharm": { "name": "#%%\n" @@ -28,110 +26,139 @@ "from utils import read_labeled_img" ] }, - { - "cell_type": "markdown", - "metadata": { - "pycharm": { - "name": "#%% md\n" - } - }, - "source": [ - "## 读取数据与构建数据集" - ] - }, { "cell_type": "code", - "execution_count": 7, - "metadata": { - "pycharm": { - "name": "#%%\n" - } - }, + "execution_count": null, "outputs": [], "source": [ - "data_dir = \"data/dataset\"\n", - "color_dict = {(0, 0, 255): \"yangeng\", (255, 0, 0): 'beijing'}\n", - "label_index = {\"yangeng\": 1, \"beijing\": 0}\n", - "dataset = read_labeled_img(data_dir, color_dict=color_dict, is_ps_color_space=False)\n", - "rus = RandomUnderSampler(random_state=0)\n", - "x_list, y_list = np.concatenate([v for k, v in dataset.items()], axis=0).tolist(), \\\n", - " np.concatenate([np.ones((v.shape[0],)) * label_index[k] for k, v in dataset.items()], axis=0).tolist()\n", + "train_from_existed = True # 是否从现有数据训练，如果是的话，那就从dataset_file训练，否则就用data_dir里头的数据\n", + "data_dir = \"data/dataset\" # 数据集，文件夹下必须包含`img`和`label`两个文件夹，放置相同文件名的图片和label\n", + "dataset_file = \"data/dataset/dataset_2022-07-20_10-04.mat\"\n", "\n", - "x_resampled, y_resampled = rus.fit_resample(x_list, y_list)\n", - "dataset = {\"inside\": np.array(x_resampled)}" - ] + "color_dict = {(0, 0, 255): \"yangeng\", (255, 0, 0): 'beijing'} # 颜色对应的类别\n", + "label_index = {\"yangeng\": 1, \"beijing\": 0} # 类别对应的序号\n", + "show_samples = False # 是否展示样本\n", + "\n", + "# 定义一些训练量\n", + "threshold = 5 # 正样本周围多大范围内的还算是正样本\n", + "node_num = 20 # 如果使用ELM作为分类器物，有多少的节点\n", + "negative_sample_num = None # None或者一个数字，对应生成的负样本数量" + ], + "metadata": { + "collapsed": false, + "pycharm": { + "name": "#%%\n" + } + } }, { "cell_type": "markdown", - "metadata": { - "pycharm": { - "name": "#%% md\n" - } - }, "source": [ - "## 模型训练" - ] + "## 读取数据" + ], + "metadata": { + "collapsed": false + } }, { "cell_type": "code", - "execution_count": 8, - "metadata": { - "pycharm": { - "name": "#%%\n" - } - }, - "outputs": [], - "source": [ - "# 定义一些常量\n", - "threshold = 5\n", - "node_num = 20\n", - "negative_sample_num = None # None或者一个数字\n", - "world_boundary = np.array([0, 0, 0, 255, 255, 255])\n", - "# 对数据进行预处理\n", - "x = np.concatenate([v for k, v in dataset.items()], axis=0)\n", - "negative_sample_num = int(x.shape[0] * 1.2) if negative_sample_num is None else negative_sample_num\n" - ] - }, - { - "cell_type": "code", - "execution_count": 2, - "metadata": { - "pycharm": { - "name": "#%%\n" - } - }, - "outputs": [], - "source": [ - "model = AnonymousColorDetector()" - ] - }, - { - "cell_type": "code", - "execution_count": 3, + "execution_count": 10, "outputs": [ { - "name": "stdout", - "output_type": "stream", - "text": [ - " precision recall f1-score support\n", - "\n", - " 0.0 0.99 0.99 0.99 26314\n", - " 1.0 0.99 0.99 0.99 24492\n", - "\n", - " accuracy 0.99 50806\n", - " macro avg 0.99 0.99 0.99 50806\n", - "weighted avg 0.99 0.99 0.99 50806\n", - "\n" + "ename": "FileNotFoundError", + "evalue": "[Errno 2] No such file or directory: 'data/dataset/label'", + "output_type": "error", + "traceback": [ + "\u001B[0;31m---------------------------------------------------------------------------\u001B[0m", + "\u001B[0;31mFileNotFoundError\u001B[0m Traceback (most recent call last)", + "\u001B[0;32m/var/folders/wh/kr5c3dr12834pfk3j7yqnrq40000gn/T/ipykernel_30867/1942905945.py\u001B[0m in \u001B[0;36m\u001B[0;34m\u001B[0m\n\u001B[0;32m----> 1\u001B[0;31m \u001B[0mdataset\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0mread_labeled_img\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mdata_dir\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mcolor_dict\u001B[0m\u001B[0;34m=\u001B[0m\u001B[0mcolor_dict\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mis_ps_color_space\u001B[0m\u001B[0;34m=\u001B[0m\u001B[0;32mFalse\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[0m\u001B[1;32m 2\u001B[0m \u001B[0;32mif\u001B[0m \u001B[0mshow_samples\u001B[0m\u001B[0;34m:\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[1;32m 3\u001B[0m \u001B[0;32mfrom\u001B[0m \u001B[0mutils\u001B[0m \u001B[0;32mimport\u001B[0m \u001B[0mlab_scatter\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[1;32m 4\u001B[0m \u001B[0mlab_scatter\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mdataset\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mclass_max_num\u001B[0m\u001B[0;34m=\u001B[0m\u001B[0;36m30000\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mis_3d\u001B[0m\u001B[0;34m=\u001B[0m\u001B[0;32mTrue\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mis_ps_color_space\u001B[0m\u001B[0;34m=\u001B[0m\u001B[0;32mFalse\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n", + "\u001B[0;32m~/PycharmProjects/tobacco_color/utils.py\u001B[0m in \u001B[0;36mread_labeled_img\u001B[0;34m(dataset_dir, color_dict, ext, is_ps_color_space)\u001B[0m\n\u001B[1;32m 37\u001B[0m \u001B[0;34m:\u001B[0m\u001B[0;32mreturn\u001B[0m\u001B[0;34m:\u001B[0m \u001B[0m字典形式的数据集\u001B[0m\u001B[0;34m{\u001B[0m\u001B[0mlabel\u001B[0m\u001B[0;34m:\u001B[0m \u001B[0mvector\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mn\u001B[0m \u001B[0mx\u001B[0m \u001B[0;36m3\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m}\u001B[0m\u001B[0;34m,\u001B[0m\u001B[0mvector为lab色彩空间\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[1;32m 38\u001B[0m \"\"\"\n\u001B[0;32m---> 39\u001B[0;31m img_names = [img_name for img_name in os.listdir(os.path.join(dataset_dir, 'label'))\n\u001B[0m\u001B[1;32m 40\u001B[0m if img_name.endswith(ext)]\n\u001B[1;32m 41\u001B[0m \u001B[0mtotal_dataset\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0mMergeDict\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n", + "\u001B[0;31mFileNotFoundError\u001B[0m: [Errno 2] No such file or directory: 'data/dataset/label'" ] } ], "source": [ - "# model.fit(x, world_boundary, threshold, negative_sample_size=negative_sample_num, train_size=0.7,\n", - "# is_save_dataset=True, model_selection='dt')\n", - "data = scipy.io.loadmat('data/dataset/dataset_2022-07-20_10-04.mat')\n", - "x, y = data['x'], data['y'].ravel()\n", - "model.fit(x, y=y, is_generate_negative=False, model_selection='dt')\n", - "model.save()" + "dataset = read_labeled_img(data_dir, color_dict=color_dict, is_ps_color_space=False)\n", + "if show_samples:\n", + " from utils import lab_scatter\n", + " lab_scatter(dataset, class_max_num=30000, is_3d=True, is_ps_color_space=False)" + ], + "metadata": { + "collapsed": false, + "pycharm": { + "name": "#%%\n" + } + } + }, + { + "cell_type": "markdown", + "source": [ + "## 数据平衡化" + ], + "metadata": { + "collapsed": false + } + }, + { + "cell_type": "code", + "execution_count": 11, + "outputs": [ + { + "ename": "NameError", + "evalue": "name 'dataset' is not defined", + "output_type": "error", + "traceback": [ + "\u001B[0;31m---------------------------------------------------------------------------\u001B[0m", + "\u001B[0;31mNameError\u001B[0m Traceback (most recent call last)", + "\u001B[0;32m/var/folders/wh/kr5c3dr12834pfk3j7yqnrq40000gn/T/ipykernel_30867/603974095.py\u001B[0m in \u001B[0;36m\u001B[0;34m\u001B[0m\n\u001B[1;32m 1\u001B[0m \u001B[0mrus\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0mRandomUnderSampler\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mrandom_state\u001B[0m\u001B[0;34m=\u001B[0m\u001B[0;36m0\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[0;32m----> 2\u001B[0;31m \u001B[0mx_list\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0my_list\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0mnp\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mconcatenate\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34m[\u001B[0m\u001B[0mv\u001B[0m \u001B[0;32mfor\u001B[0m \u001B[0mk\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mv\u001B[0m \u001B[0;32min\u001B[0m \u001B[0mdataset\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mitems\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m]\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0maxis\u001B[0m\u001B[0;34m=\u001B[0m\u001B[0;36m0\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mtolist\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m,\u001B[0m\u001B[0;31m \u001B[0m\u001B[0;31m\\\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[0m\u001B[1;32m 3\u001B[0m \u001B[0mnp\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mconcatenate\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34m[\u001B[0m\u001B[0mnp\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mones\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mv\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mshape\u001B[0m\u001B[0;34m[\u001B[0m\u001B[0;36m0\u001B[0m\u001B[0;34m]\u001B[0m\u001B[0;34m,\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m)\u001B[0m \u001B[0;34m*\u001B[0m \u001B[0mlabel_index\u001B[0m\u001B[0;34m[\u001B[0m\u001B[0mk\u001B[0m\u001B[0;34m]\u001B[0m \u001B[0;32mfor\u001B[0m \u001B[0mk\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mv\u001B[0m \u001B[0;32min\u001B[0m \u001B[0mdataset\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mitems\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m]\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0maxis\u001B[0m\u001B[0;34m=\u001B[0m\u001B[0;36m0\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mtolist\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[1;32m 4\u001B[0m \u001B[0mx_resampled\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0my_resampled\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0mrus\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mfit_resample\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mx_list\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0my_list\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[1;32m 5\u001B[0m \u001B[0mdataset\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0;34m{\u001B[0m\u001B[0;34m\"inside\"\u001B[0m\u001B[0;34m:\u001B[0m \u001B[0mnp\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0marray\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mx_resampled\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m}\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n", + "\u001B[0;31mNameError\u001B[0m: name 'dataset' is not defined" + ] + } + ], + "source": [ + "rus = RandomUnderSampler(random_state=0)\n", + "x_list, y_list = np.concatenate([v for k, v in dataset.items()], axis=0).tolist(), \\\n", + " np.concatenate([np.ones((v.shape[0],)) * label_index[k] for k, v in dataset.items()], axis=0).tolist()\n", + "x_resampled, y_resampled = rus.fit_resample(x_list, y_list)\n", + "dataset = {\"inside\": np.array(x_resampled)}" + ], + "metadata": { + "collapsed": false, + "pycharm": { + "name": "#%%\n" + } + } + }, + { + "cell_type": "markdown", + "source": [ + "## 模型训练" + ], + "metadata": { + "collapsed": false + } + }, + { + "cell_type": "code", + "execution_count": 12, + "outputs": [ + { + "ename": "NameError", + "evalue": "name 'dataset' is not defined", + "output_type": "error", + "traceback": [ + "\u001B[0;31m---------------------------------------------------------------------------\u001B[0m", + "\u001B[0;31mNameError\u001B[0m Traceback (most recent call last)", + "\u001B[0;32m/var/folders/wh/kr5c3dr12834pfk3j7yqnrq40000gn/T/ipykernel_30867/1828483636.py\u001B[0m in \u001B[0;36m\u001B[0;34m\u001B[0m\n\u001B[1;32m 1\u001B[0m \u001B[0;31m# 对数据进行预处理\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[0;32m----> 2\u001B[0;31m \u001B[0mx\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0mnp\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mconcatenate\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34m[\u001B[0m\u001B[0mv\u001B[0m \u001B[0;32mfor\u001B[0m \u001B[0mk\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mv\u001B[0m \u001B[0;32min\u001B[0m \u001B[0mdataset\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mitems\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m]\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0maxis\u001B[0m\u001B[0;34m=\u001B[0m\u001B[0;36m0\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[0m\u001B[1;32m 3\u001B[0m \u001B[0mnegative_sample_num\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0mint\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mx\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mshape\u001B[0m\u001B[0;34m[\u001B[0m\u001B[0;36m0\u001B[0m\u001B[0;34m]\u001B[0m \u001B[0;34m*\u001B[0m \u001B[0;36m1.2\u001B[0m\u001B[0;34m)\u001B[0m \u001B[0;32mif\u001B[0m \u001B[0mnegative_sample_num\u001B[0m \u001B[0;32mis\u001B[0m \u001B[0;32mNone\u001B[0m \u001B[0;32melse\u001B[0m \u001B[0mnegative_sample_num\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[1;32m 4\u001B[0m \u001B[0mmodel\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0mAnonymousColorDetector\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n", + "\u001B[0;31mNameError\u001B[0m: name 'dataset' is not defined" + ] + } + ], + "source": [ + "# 对数据进行预处理\n", + "x = np.concatenate([v for k, v in dataset.items()], axis=0)\n", + "negative_sample_num = int(x.shape[0] * 1.2) if negative_sample_num is None else negative_sample_num\n", + "model = AnonymousColorDetector()" ], "metadata": { "collapsed": false, @@ -142,9 +169,31 @@ }, { "cell_type": "code", - "execution_count": null, - "outputs": [], - "source": [], + "execution_count": 14, + "outputs": [ + { + "ename": "NameError", + "evalue": "name 'train_from_existed' is not defined", + "output_type": "error", + "traceback": [ + "\u001B[0;31m---------------------------------------------------------------------------\u001B[0m", + "\u001B[0;31mNameError\u001B[0m Traceback (most recent call last)", + "\u001B[0;32m/var/folders/wh/kr5c3dr12834pfk3j7yqnrq40000gn/T/ipykernel_30867/103680055.py\u001B[0m in \u001B[0;36m\u001B[0;34m\u001B[0m\n\u001B[0;32m----> 1\u001B[0;31m \u001B[0;32mif\u001B[0m \u001B[0mtrain_from_existed\u001B[0m\u001B[0;34m:\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[0m\u001B[1;32m 2\u001B[0m \u001B[0mdata\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0mscipy\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mio\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mloadmat\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mdataset_file\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[1;32m 3\u001B[0m \u001B[0mx\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0my\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0mdata\u001B[0m\u001B[0;34m[\u001B[0m\u001B[0;34m'x'\u001B[0m\u001B[0;34m]\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mdata\u001B[0m\u001B[0;34m[\u001B[0m\u001B[0;34m'y'\u001B[0m\u001B[0;34m]\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mravel\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[1;32m 4\u001B[0m \u001B[0mmodel\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mfit\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mx\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0my\u001B[0m\u001B[0;34m=\u001B[0m\u001B[0my\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mis_generate_negative\u001B[0m\u001B[0;34m=\u001B[0m\u001B[0;32mFalse\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mmodel_selection\u001B[0m\u001B[0;34m=\u001B[0m\u001B[0;34m'dt'\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n\u001B[1;32m 5\u001B[0m \u001B[0mmodel\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0msave\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n", + "\u001B[0;31mNameError\u001B[0m: name 'train_from_existed' is not defined" + ] + } + ], + "source": [ + "if train_from_existed:\n", + " data = scipy.io.loadmat(dataset_file)\n", + " x, y = data['x'], data['y'].ravel()\n", + " model.fit(x, y=y, is_generate_negative=False, model_selection='dt')\n", + "else:\n", + " world_boundary = np.array([0, 0, 0, 255, 255, 255])\n", + " model.fit(x, world_boundary, threshold, negative_sample_size=negative_sample_num, train_size=0.7,\n", + " is_save_dataset=True, model_selection='dt')\n", + "model.save()" + ], "metadata": { "collapsed": false, "pycharm": { diff --git a/README.md b/README.md index 2ebd9cf..0f9ceec 100644 --- a/README.md +++ b/README.md @@ -2,4 +2,68 @@ 2022年7月18日开始开发的项目，使用彩色相机进行烟梗颜色的识别。 -项目中的数据放置于[链接](https://macrosolid-my.sharepoint.com/:f:/g/personal/feijinti_miaow_fun/ElIAerke0rVMjcSyrbjetAYBpide1s-rPEY9JDZWrrhiUA?e=hrzOfa) \ No newline at end of file +## 如何进行模型训练和部署？ +1. 项目当中需要包含`data`和`models`这两个文件夹，请下载到当前文件夹下,这是链接：[data](https://macrosolid-my.sharepoint.com/personal/feijinti_miaow_fun/_layouts/15/onedrive.aspx?id=%2Fpersonal%2Ffeijinti%5Fmiaow%5Ffun%2FDocuments%2FPycharmProjects%2Ftobacco%5Fcolor%2Fdata&ga=1), [models](https://macrosolid-my.sharepoint.com/:f:/g/personal/feijinti_miaow_fun/EiyBjWEX90JGn8S-e5Kh7N8B1GWvfvDcNbpleWDTwkDm1w?e=wyL4EF) +2. 使用[01_dataset.ipynb](./01_dataset.ipynb) 进行数据集的分析文件格式需要设置为这种形式： + ```text + dataset + ├── label + │ ├── img1.bmp + │ └── ... + └── img + ├── img1.bmp + └── ... + ``` +3. 使用[02_classification.ipynb](./02_classification.ipynb)进行训练 +4. 使用[03_data_update.ipynb](02_classification.ipynb)进行数据的更新与添加 +5. 使用`main_test.py`文件进行读图测试 +6. **部署**，复制`utils.py`、`models.py`、`main.py`、`models`、`config.py`这5个文件或文件夹，运行main.py来提供预测服务。 + + +## 训练的原理 +为了应对工业环境当中负样本少的特点，我们结合颜色有限空间的特性对我们的训练过程进行了优化，核心的优化方式在于制造负样本 + +### 负样本是怎么造出来的？ + + 我们对于一个给定的色彩空间进行随机的生成一些数据，然后判断它是否是给定的正样本附近，如果是在附近，那么我们就把这些点看作是正样本，如果离得比较远，那么就会被当作是负样本。 + +### 训练的结果 + +这样子进行训练，模型就会被约束在我们给定的样本范围内，就像你看到的这样。 + +![image-20220721153751824](https://raw.githubusercontent.com/Karllzy/imagebed/main/img/image-20220721153751824.png) + +在这里，绿色就是目标的色彩范围，橙色的和蓝色则表明了模型的判定范围，模型认为蓝色的区域就是烟梗，而橙色的区域就不是烟梗。 + +可以看到，蓝色区域与绿色区域是高度重叠的，并且蓝色比绿色区域要大一些的，这正是我们想要的效果。这表明模型对于烟梗的颜色有适度的宽容，允许色彩有一定的偏差，但大体上是要达到烟梗颜色范围内的。 + +这样的好处在于，即使出现了新的杂质，只要这些杂质的色彩不在模型的宽容范围内（蓝色范围内），那么都会被判定为杂质。 + +## 预测过程的后处理（异色问题） + +### 问题的发现 + +![image-20220721154731187](https://raw.githubusercontent.com/Karllzy/imagebed/main/img/image-20220721154731187.png) + +在摄影过程中，由于相机、镜头和拍摄物体多方面的原因会出现色散边的现象，就像上图这样本来应该黄色的烟梗，边缘却变成了绿色或紫色的。 + +这是由于不同波长的光折射率不同，到达成像单元的位置会出现细小的偏差，而我们的成像单元又比较的细小，举例来说，这可能使得物体上同样的点发出的红光到了1号像素，而发出的绿光本来应该也射到1号像素却射到了相邻的2号像素，这就导致色彩不对了。 + +根据资料，一般的解决方案是对于不同波长的光进行折射率补偿，使用抗色散镜头。 + +## 镜头的影响 + +我们现有的镜头包括广角和窄角两个，这两个镜头有着不同的成像效果，如下图所示。 + +| 视角 | 广角镜头 | 窄角镜头 | +| -------- | :----------------------------------------------------------: | :----------------------------------------------------------: | +| 普通视角 | ![image-20220721155919349](https://raw.githubusercontent.com/Karllzy/imagebed/main/img/image-20220721155919349.png) | ![image-20220721155848160](https://raw.githubusercontent.com/Karllzy/imagebed/main/img/image-20220721155848160.png) | + +但是由于条件有限，我们这里就只能用算法的形式硬抗这些误差了。 + + + +## 模型的更新 + +### 如何应对新的目标物？ +