{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"import torch \nimport torchvision\nimport torch.nn as nn\nimport torch.optim as optim\nimport torchvision.transforms.functional as TF\n#-------------------------------------------\nimport os\nimport shutil\nimport zipfile\n#-------------------------------------------\nimport numpy as np\nfrom PIL import Image\nimport albumentations as A\nfrom torch.utils.data import Dataset\nfrom torch.utils.data import DataLoader\nfrom albumentations.pytorch import ToTensorV2\n#-------------------------------------------\nfrom tqdm import tqdm","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2023-10-14T13:46:33.506772Z","iopub.execute_input":"2023-10-14T13:46:33.507290Z","iopub.status.idle":"2023-10-14T13:46:42.954135Z","shell.execute_reply.started":"2023-10-14T13:46:33.507261Z","shell.execute_reply":"2023-10-14T13:46:42.953157Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
\n

Объявление путей и работа с архивами

\n

Declaring paths and working with archives

\n
\n","metadata":{"execution":{"iopub.status.busy":"2023-10-11T17:42:24.700499Z","iopub.execute_input":"2023-10-11T17:42:24.700955Z","iopub.status.idle":"2023-10-11T17:42:24.709911Z","shell.execute_reply.started":"2023-10-11T17:42:24.700923Z","shell.execute_reply":"2023-10-11T17:42:24.707788Z"}}},{"cell_type":"code","source":"path_to_data='/kaggle/input/carvana-image-masking-challenge/'","metadata":{"execution":{"iopub.status.busy":"2023-10-14T13:46:53.023220Z","iopub.execute_input":"2023-10-14T13:46:53.023674Z","iopub.status.idle":"2023-10-14T13:46:53.028190Z","shell.execute_reply.started":"2023-10-14T13:46:53.023647Z","shell.execute_reply":"2023-10-14T13:46:53.027195Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"paths=os.listdir('/kaggle/input/carvana-image-masking-challenge')","metadata":{"execution":{"iopub.status.busy":"2023-10-14T13:46:53.918853Z","iopub.execute_input":"2023-10-14T13:46:53.919490Z","iopub.status.idle":"2023-10-14T13:46:53.924182Z","shell.execute_reply.started":"2023-10-14T13:46:53.919429Z","shell.execute_reply":"2023-10-14T13:46:53.923312Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"paths","metadata":{"execution":{"iopub.status.busy":"2023-10-14T13:46:55.051671Z","iopub.execute_input":"2023-10-14T13:46:55.051994Z","iopub.status.idle":"2023-10-14T13:46:55.058809Z","shell.execute_reply.started":"2023-10-14T13:46:55.051968Z","shell.execute_reply":"2023-10-14T13:46:55.057813Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"need_list=[0,3,-3]","metadata":{"execution":{"iopub.status.busy":"2023-10-14T13:46:56.043034Z","iopub.execute_input":"2023-10-14T13:46:56.043689Z","iopub.status.idle":"2023-10-14T13:46:56.048316Z","shell.execute_reply.started":"2023-10-14T13:46:56.043659Z","shell.execute_reply":"2023-10-14T13:46:56.047198Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"base_path='/kaggle/working/'\n\nfor indx in tqdm(need_list):\n with zipfile.ZipFile(path_to_data+paths[indx], 'r') as zip_file:\n zip_file.extractall(base_path)","metadata":{"execution":{"iopub.status.busy":"2023-10-14T13:46:56.942777Z","iopub.execute_input":"2023-10-14T13:46:56.943381Z","iopub.status.idle":"2023-10-14T13:49:56.004316Z","shell.execute_reply.started":"2023-10-14T13:46:56.943352Z","shell.execute_reply":"2023-10-14T13:49:56.003421Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"\nos.mkdir('/kaggle/working/saved_imgs')","metadata":{"execution":{"iopub.status.busy":"2023-10-14T14:12:52.633518Z","iopub.execute_input":"2023-10-14T14:12:52.633845Z","iopub.status.idle":"2023-10-14T14:12:52.638933Z","shell.execute_reply.started":"2023-10-14T14:12:52.633820Z","shell.execute_reply":"2023-10-14T14:12:52.637680Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"os.mkdir('/kaggle/working/val')\nos.mkdir('/kaggle/working/val_masks')\n\nfor file in sorted(os.listdir(base_path+'train'))[4600:]:\n shutil.move('/kaggle/working/train/' + file, '/kaggle/working/val')\n\nfor file in sorted(os.listdir(base_path+'train_masks'))[4600:]:\n shutil.move('/kaggle/working/train_masks/' + file, '/kaggle/working/val_masks')","metadata":{"execution":{"iopub.status.busy":"2023-10-14T13:49:58.655929Z","iopub.execute_input":"2023-10-14T13:49:58.656230Z","iopub.status.idle":"2023-10-14T13:49:58.720004Z","shell.execute_reply.started":"2023-10-14T13:49:58.656206Z","shell.execute_reply":"2023-10-14T13:49:58.719053Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
\n

Создание функций для работы с моделью

\n

Creating functions to work with the model

\n
\n","metadata":{"execution":{"iopub.status.busy":"2023-10-11T18:18:29.548325Z","iopub.execute_input":"2023-10-11T18:18:29.548758Z","iopub.status.idle":"2023-10-11T18:18:29.556775Z","shell.execute_reply.started":"2023-10-11T18:18:29.548729Z","shell.execute_reply":"2023-10-11T18:18:29.555286Z"}}},{"cell_type":"code","source":"def save_checkpoint(state,filename='model_ch_.pth.tar'):\n print('====Saving checkpoint====')\n torch.save(state,filename)\n \n \ndef load_checkpoint(checkpoint,model):\n print('====Loading checkpoint====')\n model.load_state_dict(checkpoint[\"state_dict\"])","metadata":{"execution":{"iopub.status.busy":"2023-10-14T13:50:01.039181Z","iopub.execute_input":"2023-10-14T13:50:01.039539Z","iopub.status.idle":"2023-10-14T13:50:01.044561Z","shell.execute_reply.started":"2023-10-14T13:50:01.039511Z","shell.execute_reply":"2023-10-14T13:50:01.043416Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def get_loaders(Train_dir,\n Train_mask_dir,\n Val_img_dir,\n Val_mask_dir,\n Batch_size,\n train_transform,\n val_transform,\n Num_workers,\n PIN_MEMORY):\n train_ds = CarvanaDataset(image_dir=Train_dir,\n mask_dir=Train_mask_dir,\n transform=train_transform)\n\n train_loader = DataLoader(\n train_ds,\n batch_size=Batch_size,\n num_workers=Num_workers,\n pin_memory=PIN_MEMORY,\n shuffle=True\n )\n Val_ds = CarvanaDataset(image_dir=Val_img_dir,\n mask_dir=Val_mask_dir,\n transform=val_transform)\n\n val_loader = DataLoader(\n Val_ds,\n batch_size=Batch_size,\n num_workers=Num_workers,\n pin_memory=PIN_MEMORY,\n shuffle=False\n )\n \n\n return train_loader,val_loader\n","metadata":{"execution":{"iopub.status.busy":"2023-10-14T13:50:46.347062Z","iopub.execute_input":"2023-10-14T13:50:46.347608Z","iopub.status.idle":"2023-10-14T13:50:46.357860Z","shell.execute_reply.started":"2023-10-14T13:50:46.347570Z","shell.execute_reply":"2023-10-14T13:50:46.356687Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def check_accuracy(loader, model, device=\"cuda\"):\n num_correct=0\n num_pixels=0\n dice_score=0\n model.eval()\n \n with torch.no_grad():\n for x,y in loader:\n x=x.to(device)\n y=y.to(device)\n \n preds = torch.sigmoid(model(x))\n preds=preds.squeeze(1)\n #print(y.shape,\"yyyyy\")\n #print(preds.shape)\n preds = (preds>0.5).float()\n num_correct +=(preds==y).sum()\n \n num_pixels += torch.numel(preds)\n \n dice_score += (2*(preds*y).sum())/(\n (preds+y).sum()+1e-8\n )\n \n \n print(\n f\"Получила {num_correct}/{num_pixels} с точностью {(num_correct/num_pixels)*100:.2f}\")\n \n print(f\"Dice_score {dice_score/len(loader)}\")\n \n \n model.train()\n ","metadata":{"execution":{"iopub.status.busy":"2023-10-14T15:17:28.225381Z","iopub.execute_input":"2023-10-14T15:17:28.226047Z","iopub.status.idle":"2023-10-14T15:17:28.233015Z","shell.execute_reply.started":"2023-10-14T15:17:28.226018Z","shell.execute_reply":"2023-10-14T15:17:28.231840Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def save_prediction_as_imgs(\n loader,\n model,\n folder='/kaggle/working/',\n device='cuda'\n):\n model.eval()\n \n for idx, (x,y) in enumerate(loader):\n x=x.to(device)\n with torch.no_grad():\n preds = torch.sigmoid(model(x))\n preds = (preds>0.5).float()\n \n torchvision.utils.save_image(\n preds,f\"{folder}/pred_{idx}.png\"\n )\n \n torchvision.utils.save_image(y.unsqueeze(1),f\"{folder}/pred_{idx}.png\")\n \n \n model.train()\n \n ","metadata":{"execution":{"iopub.status.busy":"2023-10-14T14:20:32.118015Z","iopub.execute_input":"2023-10-14T14:20:32.118389Z","iopub.status.idle":"2023-10-14T14:20:32.125201Z","shell.execute_reply.started":"2023-10-14T14:20:32.118361Z","shell.execute_reply":"2023-10-14T14:20:32.124268Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
\n

Создание датасета

\n

Creating a dataset

\n
\n","metadata":{"execution":{"iopub.status.busy":"2023-10-11T17:42:39.518833Z","iopub.execute_input":"2023-10-11T17:42:39.519278Z","iopub.status.idle":"2023-10-11T17:42:39.527968Z","shell.execute_reply.started":"2023-10-11T17:42:39.519248Z","shell.execute_reply":"2023-10-11T17:42:39.526063Z"}}},{"cell_type":"code","source":"Learn_rate=1e-4\nDevice='cuda' if torch.cuda.is_available() else 'cpu'\nBatch_size=16\nNum_epoch=15\nNum_workers=2\nIMG_Height=320 #1280\nIMG_Width=480#1918\nPIN_MEMORY=True\nLoad_MOdel=False\nTrain_dir='/kaggle/working/train'\nTrain_mask_dir='/kaggle/working/train_masks'\nVal_dir='/kaggle/working/val'\nVasl_mask_dir='/kaggle/working/val_masks'\n","metadata":{"execution":{"iopub.status.busy":"2023-10-14T14:57:47.281768Z","iopub.execute_input":"2023-10-14T14:57:47.282170Z","iopub.status.idle":"2023-10-14T14:57:47.289080Z","shell.execute_reply.started":"2023-10-14T14:57:47.282139Z","shell.execute_reply":"2023-10-14T14:57:47.287807Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_transform=A.Compose(\n [\n A.Resize(height=IMG_Height,width=IMG_Width),\n A.Rotate(limit=35,p=1.0),\n A.HorizontalFlip(p=0.5),\n A.VerticalFlip(p=0.1),\n A.Normalize(\n mean=[0.0,0.0,0.0],\n std=[1.0,1.0,1.0],\n max_pixel_value=255.0\n ),\n ToTensorV2(),\n \n \n \n \n ]\n)\n\nval_transform=A.Compose(\n [\n A.Resize(height=IMG_Height,width=IMG_Width),\n A.Normalize(\n mean=[0.0,0.0,0.0],\n std=[1.0,1.0,1.0],\n max_pixel_value=255.0\n ),\n ToTensorV2(),\n \n ]\n)\n\n\n","metadata":{"execution":{"iopub.status.busy":"2023-10-14T13:50:55.044055Z","iopub.execute_input":"2023-10-14T13:50:55.045135Z","iopub.status.idle":"2023-10-14T13:50:55.052308Z","shell.execute_reply.started":"2023-10-14T13:50:55.045097Z","shell.execute_reply":"2023-10-14T13:50:55.051340Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"class CarvanaDataset(Dataset):\n def __init__(self, image_dir,mask_dir,transform=None):\n self.image_dir=image_dir\n self.mask_dir=mask_dir\n self.transform=transform\n \n self.images=os.listdir(image_dir)\n#================================================================================== \n \n \n def __len__(self):\n return len(self.images)\n \n#================================================================================== \n def __getitem__(self,indx):\n \n img_path=os.path.join(self.image_dir,self.images[indx])\n mask_path=os.path.join(self.mask_dir,self.images[indx].replace('.jpg','_mask.gif'))\n \n \n image=np.array(Image.open(img_path).convert('RGB'))\n \n mask=np.array(Image.open(mask_path).convert('L'),dtype=np.float32)\n \n #mask[mask_path==255.0]=1.0\n mask=mask/255.0\n #self.transform is not None\n if self.transform:\n augmentations=self.transform(image=image,mask=mask)\n image=augmentations['image']\n \n mask=augmentations['mask']\n \n \n return image,mask\n \n \n ","metadata":{"execution":{"iopub.status.busy":"2023-10-14T13:50:58.205139Z","iopub.execute_input":"2023-10-14T13:50:58.205497Z","iopub.status.idle":"2023-10-14T13:50:58.212465Z","shell.execute_reply.started":"2023-10-14T13:50:58.205463Z","shell.execute_reply":"2023-10-14T13:50:58.211368Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import matplotlib.pyplot as plt\ndataset=CarvanaDataset(\"/kaggle/working/val\",\"/kaggle/working/val_masks\")\n\n# Получение тензора из датасета\nimage, label = dataset[1] # Предположим, что у вас есть датасет с именем dataset и индексом idx\n\n# Преобразование тензора обратно в изображение\nimage = TF.to_pil_image(image)\n\n# Отображение изображения\nplt.imshow(image)\nplt.axis('off')\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2023-10-14T15:14:32.564376Z","iopub.execute_input":"2023-10-14T15:14:32.564752Z","iopub.status.idle":"2023-10-14T15:14:33.054325Z","shell.execute_reply.started":"2023-10-14T15:14:32.564725Z","shell.execute_reply":"2023-10-14T15:14:33.053451Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
\n

Создание модели

\n

Creating a Model

\n
\n","metadata":{"execution":{"iopub.status.busy":"2023-10-11T17:39:49.069727Z","iopub.execute_input":"2023-10-11T17:39:49.070400Z","iopub.status.idle":"2023-10-11T17:39:49.078871Z","shell.execute_reply.started":"2023-10-11T17:39:49.070367Z","shell.execute_reply":"2023-10-11T17:39:49.077078Z"}}},{"cell_type":"code","source":"class ConvBlock(nn.Module):\n def __init__(self,in_channels,out_channels):\n super(ConvBlock,self).__init__()\n self.conv=nn.Sequential(\n nn.Conv2d(in_channels,out_channels,3,1,1,bias=False),\n nn.BatchNorm2d(out_channels),\n nn.ReLU(inplace=True),\n nn.Conv2d(out_channels,out_channels,3,1,1,bias=False),\n nn.BatchNorm2d(out_channels),\n nn.ReLU(inplace=True)\n\n )\n def forward(self,x):\n return self.conv(x)\n ","metadata":{"execution":{"iopub.status.busy":"2023-10-14T13:50:59.975323Z","iopub.execute_input":"2023-10-14T13:50:59.975981Z","iopub.status.idle":"2023-10-14T13:50:59.981607Z","shell.execute_reply.started":"2023-10-14T13:50:59.975950Z","shell.execute_reply":"2023-10-14T13:50:59.980567Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"class UNET(nn.Module):\n def __init__(self,\n in_channels=3,\n out_channels=1,\n features=[64,128,256,512]):\n \n super(UNET,self).__init__()\n self.uppers=nn.ModuleList()\n self.downs=nn.ModuleList()\n self.pool=nn.MaxPool2d(kernel_size=2,stride=2)\n \n #Нижняя часть сети UNET\n \n for feature in features:\n self.downs.append(ConvBlock(in_channels,\n feature))\n \n in_channels=feature\n \n \n #Верх сетки UNET\n for feature in reversed(features):\n self.uppers.append(\n nn.ConvTranspose2d(feature*2,\n feature,\n kernel_size=2,\n stride=2\n )\n )\n self.uppers.append(ConvBlock(feature*2,\n feature))\n \n \n self.bottleneck=ConvBlock(features[-1],\n features[-1]*2)\n \n \n \n self.last_conv=nn.Conv2d(features[0],out_channels,kernel_size=1)\n \n #=====================================================================================\n def forward(self,x):\n \n skip_connections=[]\n \n for down in self.downs:\n x= down(x)\n skip_connections.append(x)\n x= self.pool(x)\n \n x= self.bottleneck(x)\n skip_connections= skip_connections[::-1]\n for idx in range(0,len(self.uppers),2):\n \n x=self.uppers[idx](x)\n skip_connection= skip_connections[idx//2]\n \n if x.shape != skip_connection.shape:\n x= TF.resize(x,size=skip_connection[2:])\n \n concat_x= torch.cat((skip_connection,x),dim=1)\n x= self.uppers[idx+1](concat_x)\n \n \n return self.last_conv(x)\n \n \n ","metadata":{"execution":{"iopub.status.busy":"2023-10-14T13:51:01.225158Z","iopub.execute_input":"2023-10-14T13:51:01.225817Z","iopub.status.idle":"2023-10-14T13:51:01.235731Z","shell.execute_reply.started":"2023-10-14T13:51:01.225786Z","shell.execute_reply":"2023-10-14T13:51:01.234508Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def test():\n x=torch.randn((3,1,224,224))\n model=UNET( in_channels=1,\n out_channels=1)\n preds=model(x)\n print(preds.shape)\n print(x.shape)\n \n assert preds.shape==x.shape","metadata":{"execution":{"iopub.status.busy":"2023-10-14T13:51:02.422376Z","iopub.execute_input":"2023-10-14T13:51:02.422741Z","iopub.status.idle":"2023-10-14T13:51:02.427670Z","shell.execute_reply.started":"2023-10-14T13:51:02.422715Z","shell.execute_reply":"2023-10-14T13:51:02.426614Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"test()","metadata":{"execution":{"iopub.status.busy":"2023-10-12T17:44:00.080906Z","iopub.execute_input":"2023-10-12T17:44:00.082150Z","iopub.status.idle":"2023-10-12T17:44:05.906032Z","shell.execute_reply.started":"2023-10-12T17:44:00.082102Z","shell.execute_reply":"2023-10-12T17:44:05.904846Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
\n

Обучение на данных

\n

Learning from data

\n
\n","metadata":{}},{"cell_type":"code","source":"","metadata":{"execution":{"iopub.status.busy":"2023-10-11T17:51:01.026952Z","iopub.execute_input":"2023-10-11T17:51:01.027354Z","iopub.status.idle":"2023-10-11T17:51:01.032931Z","shell.execute_reply.started":"2023-10-11T17:51:01.027327Z","shell.execute_reply":"2023-10-11T17:51:01.032020Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def train_(loader,model,optimizer,loss_fn,scaler):\n loop=tqdm(loader)\n \n for batch_indx,(data,targets) in enumerate(loop):\n data=data.to(device=Device)\n targets=targets.float().unsqueeze(1).to(device=Device)\n \n \n \n #train\n with torch.cuda.amp.autocast():\n \n pred=model(data)\n loss=loss_fn(pred,targets)\n \n \n \n optimizer.zero_grad()\n scaler.scale(loss).backward()\n scaler.step(optimizer)\n \n scaler.update()\n \n loop.set_postfix(loss=loss.item())\n \n \n \n \n ","metadata":{"execution":{"iopub.status.busy":"2023-10-14T13:53:40.717802Z","iopub.execute_input":"2023-10-14T13:53:40.718156Z","iopub.status.idle":"2023-10-14T13:53:40.725538Z","shell.execute_reply.started":"2023-10-14T13:53:40.718125Z","shell.execute_reply":"2023-10-14T13:53:40.724282Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"
\n

Оптимизация Работы Видеокарты

\n

В данной работе особое внимание уделяется использованию scaler для оптимизации работы видеокарты. Эта возможность играет ключевую роль при переносе весов модели на ПЛИС и может оказаться весьма полезной в контексте оптимизации производительности.

\n

Optimizing GPU Performance

\n

A crucial aspect of this work is the utilization of scaler to optimize GPU performance. This feature becomes especially valuable when transferring model weights to FPGA, offering significant advantages in performance optimization.

\n
","metadata":{}},{"cell_type":"code","source":"model=UNET(in_channels=3,out_channels=1).to(device=Device)\nloss_fn=nn.BCEWithLogitsLoss()\noptimizer=optim.Adam(model.parameters(),lr=Learn_rate)\n\ntrain_loader,val_loader=get_loaders(\n Train_dir,\n Train_mask_dir,\n Val_dir,\n Vasl_mask_dir, \n Batch_size,\n train_transform,\n val_transform,\n Num_workers,\n PIN_MEMORY\n \n \n)\n\n\nscaler=torch.cuda.amp.GradScaler()\n\nfor epoch in range(Num_epoch):\n train_(train_loader,model,optimizer,loss_fn,scaler)\n #save model\n checkpoint={\n \"state_dict\":model.state_dict(),\n \"optimizer\":optimizer.state_dict(),\n \n }\n \n save_checkpoint(checkpoint)\n #check acc\n check_accuracy(val_loader,model,device=Device)\n #print some exmpls\n save_prediction_as_imgs(val_loader,model,folder=\"saved_imgs/\",device=Device)\n \n \n \n\n","metadata":{"execution":{"iopub.status.busy":"2023-10-14T15:17:33.132537Z","iopub.execute_input":"2023-10-14T15:17:33.132863Z","iopub.status.idle":"2023-10-14T15:59:48.467926Z","shell.execute_reply.started":"2023-10-14T15:17:33.132838Z","shell.execute_reply":"2023-10-14T15:59:48.466707Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"code","source":" torch.save(model.state_dict(), f\"pytorch_model-e{Num_epoch}.pth\")","metadata":{"execution":{"iopub.status.busy":"2023-10-14T16:09:25.516543Z","iopub.execute_input":"2023-10-14T16:09:25.517759Z","iopub.status.idle":"2023-10-14T16:09:25.804535Z","shell.execute_reply.started":"2023-10-14T16:09:25.517702Z","shell.execute_reply":"2023-10-14T16:09:25.803324Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"Visual_testing_datas=CarvanaDataset(\"/kaggle/working/val\",\"/kaggle/working/val_masks\",transform=val_transform)","metadata":{"execution":{"iopub.status.busy":"2023-10-14T16:20:30.063280Z","iopub.execute_input":"2023-10-14T16:20:30.063646Z","iopub.status.idle":"2023-10-14T16:20:30.069499Z","shell.execute_reply.started":"2023-10-14T16:20:30.063617Z","shell.execute_reply":"2023-10-14T16:20:30.068227Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"Visual_testing_datas[0][0].shape","metadata":{"execution":{"iopub.status.busy":"2023-10-14T16:20:42.066978Z","iopub.execute_input":"2023-10-14T16:20:42.067363Z","iopub.status.idle":"2023-10-14T16:20:42.131582Z","shell.execute_reply.started":"2023-10-14T16:20:42.067334Z","shell.execute_reply":"2023-10-14T16:20:42.130714Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import random\nindex = os.listdir('/kaggle/working/val')\nimage_i = random.choice(index)\n\nnew_image_path ='/kaggle/working/val/'+ image_i\nnew_image = np.array(Image.open(new_image_path).convert('RGB'))\nprint(new_image_path)\n\n\n\ninput_image = val_transform(image=new_image) # C альбументацией преобразовываем изображение\n\n# Добавил размерность, чтобы был батч\ninput_image = input_image['image'].unsqueeze(0) .to(Device)\n\nwith torch.no_grad():\n predicted_mask = model(input_image)\n\n#Использование пороговой функции\nthreshold = 0.5\npredicted_mask_binary = (predicted_mask > threshold).float()\n\n\n","metadata":{"execution":{"iopub.status.busy":"2023-10-14T17:03:11.691056Z","iopub.execute_input":"2023-10-14T17:03:11.691568Z","iopub.status.idle":"2023-10-14T17:03:11.723266Z","shell.execute_reply.started":"2023-10-14T17:03:11.691522Z","shell.execute_reply":"2023-10-14T17:03:11.722220Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(predicted_mask_binary.shape)","metadata":{"execution":{"iopub.status.busy":"2023-10-14T17:03:33.632012Z","iopub.execute_input":"2023-10-14T17:03:33.633089Z","iopub.status.idle":"2023-10-14T17:03:33.638373Z","shell.execute_reply.started":"2023-10-14T17:03:33.633048Z","shell.execute_reply":"2023-10-14T17:03:33.637410Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"predicted_mask_binary_np=np.array(predicted_mask_binary.squeeze(0).cpu())\npredicted_mask_binary_np=predicted_mask_binary_np.squeeze(0)","metadata":{"execution":{"iopub.status.busy":"2023-10-14T17:05:15.047490Z","iopub.execute_input":"2023-10-14T17:05:15.047832Z","iopub.status.idle":"2023-10-14T17:05:15.053504Z","shell.execute_reply.started":"2023-10-14T17:05:15.047806Z","shell.execute_reply":"2023-10-14T17:05:15.052410Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(predicted_mask_binary_np.shape)","metadata":{"execution":{"iopub.status.busy":"2023-10-14T17:05:21.025862Z","iopub.execute_input":"2023-10-14T17:05:21.026208Z","iopub.status.idle":"2023-10-14T17:05:21.031510Z","shell.execute_reply.started":"2023-10-14T17:05:21.026179Z","shell.execute_reply":"2023-10-14T17:05:21.030463Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"predicted_mask_array_3ch = np.repeat(predicted_mask_binary_np[:, :, np.newaxis], 3, axis=2)\npredicted_mask_tensor = torch.tensor(predicted_mask_array_3ch).permute(2, 0, 1)\ninput_image=input_image.cpu().squeeze(0)","metadata":{"execution":{"iopub.status.busy":"2023-10-14T17:05:39.200147Z","iopub.execute_input":"2023-10-14T17:05:39.200535Z","iopub.status.idle":"2023-10-14T17:05:39.207010Z","shell.execute_reply.started":"2023-10-14T17:05:39.200504Z","shell.execute_reply":"2023-10-14T17:05:39.205589Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"print(predicted_mask_tensor.shape)\nprint(input_image.shape)","metadata":{"execution":{"iopub.status.busy":"2023-10-14T17:05:44.216102Z","iopub.execute_input":"2023-10-14T17:05:44.216538Z","iopub.status.idle":"2023-10-14T17:05:44.221845Z","shell.execute_reply.started":"2023-10-14T17:05:44.216497Z","shell.execute_reply":"2023-10-14T17:05:44.220691Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"segmented_image = ((input_image + (predicted_mask_tensor * 0.5)).permute(1, 2, 0)).numpy()\npredicted_mask_tensor=predicted_mask_tensor.permute(1, 2, 0).numpy()","metadata":{"execution":{"iopub.status.busy":"2023-10-14T17:06:15.403883Z","iopub.execute_input":"2023-10-14T17:06:15.404231Z","iopub.status.idle":"2023-10-14T17:06:15.409856Z","shell.execute_reply.started":"2023-10-14T17:06:15.404196Z","shell.execute_reply":"2023-10-14T17:06:15.408909Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Визуализация оригинального изображения и сегментированного изображения\nfig, axes = plt.subplots(1, 2, figsize=(12, 6))\naxes[0].imshow(predicted_mask_tensor)\naxes[0].set_title('Оригинальное изображение/original')\naxes[0].axis('off')\n\naxes[1].imshow(segmented_image)\naxes[1].set_title('Сегментированное изображение/original with mask')\naxes[1].axis('off')\n\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2023-10-14T17:06:41.411972Z","iopub.execute_input":"2023-10-14T17:06:41.412309Z","iopub.status.idle":"2023-10-14T17:06:41.755597Z","shell.execute_reply.started":"2023-10-14T17:06:41.412281Z","shell.execute_reply":"2023-10-14T17:06:41.754674Z"},"trusted":true},"execution_count":null,"outputs":[]}]}