使用Python一步一步搭建一个深度学习GUI【不定时更新】

多巴胺耐受

已于 2024-05-31 16:27:50 修改

阅读量625

点赞数 8

分类专栏：深度学习 Python 文章标签： python 深度学习开发语言

于 2024-01-25 18:06:04 首次发布

本文链接：https://blog.csdn.net/yoto_jo/article/details/135849967

版权

本文逐步教你如何使用Python构建深度学习的图形用户界面，涵盖图像分类、语义分割等任务，涉及DeepLab、ResNet等多个模型，并讨论界面设计与皮肤样式更换。目前遇到的问题是控制台输出和进度条无法在界面上完美呈现。

摘要由CSDN通过智能技术生成

本文的网络设计大多数来自于 https://github.com/bubbliiiing?tab=repositories，截取其中关键部分，将训练的脚本设计成可视化的操作界面。
第一步是训练已有的网络结构，包含图像分类、语义分割、目标检测和实例分割。
第二步的话，我觉得是可以加载单独的torch模块进行训练或者自定义的网络结构之类的，但是还不知道怎么做，继续学习吧！
第三步，关于预测功能的实现，应该可以包含图像导入处理、视频流处理、作为服务端接收图像将结果返回客户端等功能。

目标：

包含的任务类型和网络框架：

2024-01-25 实现了语义分割DeepLab、UNet、FCN、PSPNet、SegNet模型的训练。添加HRNet和DANet选项，但是还没有实现。
2024-01-29 实现了图像分类AlexNet、DenseNet、GhostsNet、GoogleNet、MobileNet、ResNet、Shufflenet、Swin_Transformer、VGGNet、Vision_Transformer模型多种版本的训练，仅测试了AlexNet可顺利训练，其余的还没有测试。
保留了Inception、SqueezeNet和EfficientNet选项，但是未实现。
2024-02-07 测试了DenseNet、GhostsNet、GoogleNet、MobileNet都可以正常训练，但是GoogleNet无法实现冻结训练。
2024-05-31 重新调整了一下界面
在这里插入图片描述

 NetworkTypeParams = [
        {
   
            "ID": "图像分类",
            "Ranges": [
                {
   "NetworkType": "AlexNet", "Version": ["none"], "Backbones": ["none"]},
                {
   "NetworkType": "DenseNet", "Version": ["16", "19"], "Backbones": ["none"]},
                {
   "NetworkType": "GhostsNet", "Version": ["16", "19"], "Backbones": ["none"]},
                {
   "NetworkType": "GoogleNet", "Version": ["16", "19"], "Backbones": ["none"]},
                {
   "NetworkType": "MobileNet", "Version": ["v2", "v3"], "Backbones": ["none"]},
                {
   "NetworkType": "ResNet", "Version": ["18", "34", "50", "101", "152"], "Backbones": ["none"]},
                {
   "NetworkType": "Shufflenet", "Version": ["g1", "g2", "v2_x0_5", "v2_x1_0", "v2_x1_5", "v2_x2_0"], "Backbones": ["none"]},
                {
   "NetworkType": "Swin_Transformer", "Version": ["tiny", "small", "base"], "Backbones": ["none"]},
                {
   "NetworkType": "VGGNet", "Version": ["11", "13", "16", "19", "11_bn", "13_bn", "16_bn", "19_bn"], "Backbones": ["none"]},
                {
   "NetworkType": "Vision_Transformer", "Version": ["vit_b_16"], "Backbones": ["none"]},

                {
   "NetworkType": "Inception", "Version": ["1.0", "2.0"], "Backbones": ["none"]},
                {
   "NetworkType": "SqueezeNet", "Version": ["1.0", "1.1"], "Backbones": ["none"]},
                {
   "NetworkType": "EfficientNet", "Version": ["B0", "B1", "B2", "B3", "B4", "B5", "B6", "B7"], "Backbones": ["none"]}
            ],
            "默认值": {
   "NetworkType": "AlexNet", "Version": "none", "Backbones": ["none"]}
        },
        {
   
            "ID": "目标检测",
            "Ranges": [
                {
   "NetworkType": "YOLO", "Version": ["v1", "v2", "v3", "v4", "v5", "v6", "v7", "v8"], "Backbones": ["none"]},
                {
   "NetworkType": "Faster R-CNN", "Version": ["1.0"], "Backbones": ["none"]},
                {
   "NetworkType": "SSD", "Version": ["1.0"], "Backbones": ["none"]},
                {
   "NetworkType": "RetinaNet", "Version": ["1.0"], "Backbones": ["none"]},
                {
   "NetworkType": "R-FCN", "Version": ["1.0"], "Backbones": ["none"]},
                {
   "NetworkType": "CornerNet", "Version": ["1.0"], "Backbones": ["none"]},
                {
   "NetworkType": "CenterNet", "Version": ["1.0"], "Backbones": ["none"]},
                {
   "NetworkType"