分割模型
实例分割(轴对齐矩形)
1.将图像的特征对象及逆行像素精确度的标注分类
2.调用hdl可得矩形框(轴对齐矩形),类别区域的掩膜,类别区域掩膜的概率,类别,置信度
dev_get_window (WindowHandle)
*读取神经网络模型
read_dl_model ('C:/Users/zhang/Desktop/MVTecDLT/hdl调用/实例分割轴对齐边界框混合药片.hdl', DLModelHandle)
*设置模型参数
set_dl_model_param (DLModelHandle, 'runtime', 'cpu')
*获取模型参数
get_dl_model_param(DLModelHandle, 'image_dimensions', ImageDimensions) //获取图像的尺寸224*224*3通道
get_dl_model_param(DLModelHandle, 'class_ids', ClassIDs) //获取ClassIDs,一共5类
get_dl_model_param (DLModelHandle, 'class_names', ClassNames)//获取类别的名字
*从模型里面获取图像前处理参数到字典里
create_dl_preprocess_param_from_model (DLModelHandle, 'none', 'full_domain', [], [], [], DLPreprocessParam)
*读取图像
read_image (Image, 'C:/Users/zhang/Desktop/MVTecDLT/hdl调用/pill_bag/pill_bag_001.png')
*用图像生成一个样本字典
gen_dl_samples_from_images (Image, DLSample)
*对图像的样本字典进行前处理
preprocess_dl_samples (DLSample, DLPreprocessParam)
*对处理后的样本字典送进模型进行推理,得到推理结果
apply_dl_model (DLModelHandle, DLSample, [], DLResult)
* 获取推理数据
get_dict_object (Object, DLResult, 'mask_probs')
get_dict_object (Object1, DLResult, 'mask')
get_dict_tuple (DLResult, 'bbox_row1', bbox_row1)
get_dict_tuple (DLResult, 'bbox_col1', bbox_col1)
get_dict_tuple (DLResult, 'bbox_row2', bbox_row2)
get_dict_tuple (DLResult, 'bbox_col2', bbox_col2)
get_dict_tuple (DLResult, 'bbox_class_id', bbox_class_id)
get_dict_tuple (DLResult, 'bbox_class_name', bbox_class_name)
get_dict_tuple (DLResult, 'bbox_confidence', bbox_confidence)
*dev_set_draw ('margin')
zoom_image_size (Image, ImageZoom, ImageDimensions[0], ImageDimensions[1], 'constant')
gen_rectangle1 (Rectangle, bbox_row1, bbox_col1, bbox_row2, bbox_col2)