SketchCode是一种深度学习模型，它采用手绘的Web原型并将其转换为有效的HTML代码，使用图像字幕体系结构从手绘网站线框生成HTML标记。
注意：此项目是概念验证；该模型性能依赖于类似于核心数据集的线框。
依赖
Pytho3(otcompatiblewithpytho2)pip安装依赖
pipistall-rrequiremets.txt用法示例下载数据和预训练的权重：#Gettigthedata,1,700images,342mbgitcloehttps://github.com/ashkumar/sketch-code.gitcdsketch-codecdscripts#Getthedataadpretraiedweightsshget_data.shshget_pretraied_model.sh使用预训练的权重将示例绘制的图像转换为HTML代码：cdsrcpythocovert_sigle_image.py--pg_path../examples/draw_example1.pg\--output_folder./geerated_html\--model_jso_file../bi/model_jso.jso\--model_weights_file../bi/weights.h5一般用法使用权重将单个图像转换为HTML代码：cdsrcpythocovert_sigle_image.py--pg_path{path/to/img.pg}\--output_folder{folder/to/output/html}\--model_jso_file{path/to/model/jso_file.jso}\--model_weights_file{path/to/model/weights.h5}将文件夹中的一批图像转换为HTML：cdsrcpythocovert_batch_of_images.py--pgs_path{path/to/folder/with/pgs}\--output_folder{folder/to/output/html}\--model_jso_file{path/to/model/jso_file.jso}\--model_weights_file{path/to/model/weights.h5}训练模型：cdsrc#traiigfromscratch#<augmet_traiig_data>addsKerasImageDataGeeratoraugmetatiofortraiigimagespythotrai.py--data_iput_path{path/to/folder/with/pgs/guis}\--validatio_split0.2\--epochs10\--model_output_path{path/to/output/model}--augmet_traiig_data1#traiigstartigwithpretraiedmodelpythotrai.py--data_iput_path{path/to/folder/with/pgs/guis}\--validatio_split0.2\--epochs10\--model_output_path{path/to/output/model}\--model_jso_file../bi/model_jso.jso\--model_weights_file../bi/pretraied_weights.h5\--augmet_traiig_data1使用BLEU分数评估生成的预测cdsrc#evaluatesigleGUIpredictiopythoevaluate_sigle_gui.py--origial_gui_filepath{path/to/origial/gui/file}\--predicted_gui_filepath{path/to/predicted/gui/file}#traiigstartigwithpretraiedmodelpythoevaluate_batch_guis.py--origial_guis_filepath{path/to/folder/with/origial/guis}\--predicted_guis_filepath{path/to/folder/with/predicted/guis}
评论