uploaded 15 files

app.py

@@ -0,0 +1,126 @@
+# gradioMisClassGradCAMimageInputter
+import os
+import math
+import numpy as np
+import pandas as pd
+import seaborn as sn
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import torchvision
+import matplotlib.pyplot as plt
+from pl_bolts.datamodules import CIFAR10DataModule
+from pl_bolts.transforms.dataset_normalizations import cifar10_normalization
+from pytorch_lightning import LightningModule, Trainer, seed_everything
+from pytorch_lightning.callbacks import LearningRateMonitor
+from pytorch_lightning.callbacks.progress import TQDMProgressBar
+from pytorch_lightning.loggers import CSVLogger
+from torch.optim.lr_scheduler import OneCycleLR
+from torch.optim.swa_utils import AveragedModel, update_bn
+from torchmetrics.functional import accuracy
+from pytorch_lightning.callbacks import ModelCheckpoint
+from torchvision import datasets, transforms, utils
+from PIL import Image
+from pytorch_grad_cam import GradCAM
+from pytorch_grad_cam.utils.image import show_cam_on_image
+import gradio as gr
+import misclas_helper
+import gradcam_helper
+import lightningmodel
+from misclas_helper import display_cifar_misclassified_data
+from gradcam_helper import display_gradcam_output
+from misclas_helper import get_misclassified_data2
+from lightningmodel import LitResnet
+
+fileName = None
+
+targets = None
+device = torch.device("cpu")
+classes = ('plane', 'car', 'bird', 'cat', 'deer',
+           'dog', 'frog', 'horse', 'ship', 'truck')
+
+model = LitResnet(lr=0.05).load_from_checkpoint("weights_92.ckpt")
+
+device = torch.device("cpu")
+
+# Denormalize the data using test mean and std deviation
+inv_normalize = transforms.Normalize(
+    mean=[-0.50/0.23, -0.50/0.23, -0.50/0.23],
+    std=[1/0.23, 1/0.23, 1/0.23]
+)
+
+# Get the misclassified data from test dataset
+misclassified_data = get_misclassified_data2(model, device, 20)
+
+def hello(DoYouWantToShowMisClassifiedImages, HowManyImages):
+  if(DoYouWantToShowMisClassifiedImages.lower() == "yes"):
+    fileName = misclas_helper.display_cifar_misclassified_data(misclassified_data, classes, inv_normalize, number_of_samples=HowManyImages)
+    return Image.open(fileName)
+  else:
+    return None
+misClass_demo = gr.Interface(
+    fn = hello,
+    inputs=['text', gr.Slider(0, 20, step=5)],
+    outputs=['image'],
+    title="Misclasseified Images",
+    description="If your answer to the question DoYouWantToShowMisClassifiedImages is yes, then only it works.",
+)
+
+
+############
+
+
+def inference(DoYouWantToShowGradCAMMedImages, HowManyImages, WhichLayer, transparency):
+  if(DoYouWantToShowGradCAMMedImages.lower() == "yes"):
+    if(WhichLayer == -1):
+      target_layers = [model.model.resNetLayer2Part2[-1]]
+    elif(WhichLayer == -2):
+      target_layers = [model.model.resNetLayer2Part1[-1]]
+    elif(WhichLayer == -3):
+      target_layers = [model.model.Layer3[-1]]
+    fileName = gradcam_helper.display_gradcam_output(misclassified_data, classes, inv_normalize, model.model, target_layers, targets, number_of_samples=HowManyImages, transparency=0.70)
+    return Image.open(fileName)
+
+gradCAM_demo = gr.Interface(
+    fn=inference,
+        #DoYouWantToShowGradCAMMedImages, HowManyImages, WhichLayer, transparency
+    inputs=['text', gr.Slider(0, 20, step=5), gr.Slider(-3, -1, value = -1, step=1), gr.Slider(0, 1, value = 0.7, label = "Overall Opacity of the Overlay")],
+    outputs=['image'],
+    title="GradCammd Images",
+    description="If your answer to the question DoYouWantToShowGradCAMMedImages is yes, then only it works.",
+)
+
+
+############
+
+def ImageInputter(img1, img2, img3, img4, img5, img6, img7, img8, img9, img10):
+  return img1, img2, img3, img4, img5, img6, img7, img8, img9, img10
+
+imageInputter_demo = gr.Interface(
+    ImageInputter,
+    [
+        "image","image","image","image","image","image","image","image","image","image"
+    ],
+    [
+        "image","image","image","image","image","image","image","image","image","image"
+    ],
+    examples=[
+        ["bird.jpg", "car.jpg", "cat.jpg"],
+        ["deer.jpg", "dog.jpg", "frog.jpg"],
+        ["horse.jpg", "plane.jpg", "ship.jpg"],
+        [None, "truck.jpg", None],
+    ],
+    title="Max 10 images input",
+    description="Here's a sample image inputter. Allows you to feed in 10 images and display them. You may drag and drop images from bottom examples to the input feeders",
+)
+
+
+############
+
+
+demo = gr.TabbedInterface(
+    interface_list = [misClass_demo, gradCAM_demo, imageInputter_demo],
+    tab_names = ["MisClassified Images", "GradCAMMed Images", "10 images inputter"]
+)
+
+demo.launch(debug=True)

gradcam_helper.py

@@ -0,0 +1,86 @@
+import os
+import math
+import numpy as np
+import pandas as pd
+import seaborn as sn
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import torchvision
+import matplotlib.pyplot as plt
+import torch.nn as nn
+import torch.nn.functional as F
+from pl_bolts.datamodules import CIFAR10DataModule
+from pl_bolts.transforms.dataset_normalizations import cifar10_normalization
+from pytorch_lightning import LightningModule, Trainer, seed_everything
+from pytorch_lightning.callbacks import LearningRateMonitor
+from pytorch_lightning.callbacks.progress import TQDMProgressBar
+from pytorch_lightning.loggers import CSVLogger
+from torch.optim.lr_scheduler import OneCycleLR
+from torch.optim.swa_utils import AveragedModel, update_bn
+from torchmetrics.functional import accuracy
+from pytorch_lightning.callbacks import ModelCheckpoint
+from torchvision import datasets, transforms, utils
+from PIL import Image
+from pytorch_grad_cam import GradCAM
+from pytorch_grad_cam.utils.image import show_cam_on_image
+
+
+
+targets = None
+
+
+# Yes - This is important predecessor3 for gradioMisClassGradCAM
+def display_gradcam_output(data: list,
+                           classes: list[str],
+                           inv_normalize: transforms.Normalize,
+                           model: 'DL Model',
+                           target_layers: list['model_layer'],
+                           targets=None,
+                           number_of_samples: int = 10,
+                           transparency: float = 0.60):
+    """
+    Function to visualize GradCam output on the data
+    :param data: List[Tuple(image, label)]
+    :param classes: Name of classes in the dataset
+    :param inv_normalize: Mean and Standard deviation values of the dataset
+    :param model: Model architecture
+    :param target_layers: Layers on which GradCam should be executed
+    :param targets: Classes to be focused on for GradCam
+    :param number_of_samples: Number of images to print
+    :param transparency: Weight of Normal image when mixed with activations
+    """
+    # Plot configuration
+    fig = plt.figure(figsize=(10, 10))
+    x_count = 5
+    y_count = 1 if number_of_samples <= 5 else math.floor(number_of_samples / x_count)
+
+    # Create an object for GradCam
+    #cam = GradCAM(model=model, target_layers=target_layers, use_cuda=True)
+    cam = GradCAM(model=model, target_layers=target_layers)
+
+    # Iterate over number of specified images
+    for i in range(number_of_samples):
+        plt.subplot(y_count, x_count, i + 1)
+        input_tensor = data[i][0]
+
+        # Get the activations of the layer for the images
+        grayscale_cam = cam(input_tensor=input_tensor, targets=targets)
+        grayscale_cam = grayscale_cam[0, :]
+
+        # Get back the original image
+        img = input_tensor.squeeze(0).to('cpu')
+        img = inv_normalize(img)
+        rgb_img = np.transpose(img, (1, 2, 0))
+        rgb_img = rgb_img.numpy()
+
+        # Mix the activations on the original image
+        visualization = show_cam_on_image(rgb_img, grayscale_cam, use_rgb=True, image_weight=transparency)
+
+        # Display the images on the plot
+        plt.imshow(visualization)
+        # plt.title(r"Correct: " + classes[data[i][1].item()] + '\n' + 'Output: ' + classes[data[i][2].item()])
+        plt.xticks([])
+        plt.yticks([])
+    plt.savefig('imshow_output_gradcam.png')
+    return 'imshow_output_gradcam.png'

lightningmodel.py

@@ -0,0 +1,262 @@
+import os
+import math
+import numpy as np
+import pandas as pd
+import seaborn as sn
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import torchvision
+import matplotlib.pyplot as plt
+import torch.nn as nn
+import torch.nn.functional as F
+from pl_bolts.datamodules import CIFAR10DataModule
+from pl_bolts.transforms.dataset_normalizations import cifar10_normalization
+from pytorch_lightning import LightningModule, Trainer, seed_everything
+from pytorch_lightning.callbacks import LearningRateMonitor
+from pytorch_lightning.callbacks.progress import TQDMProgressBar
+from pytorch_lightning.loggers import CSVLogger
+from torch.optim.lr_scheduler import OneCycleLR
+from torch.optim.swa_utils import AveragedModel, update_bn
+from torchmetrics.functional import accuracy
+from pytorch_lightning.callbacks import ModelCheckpoint
+from torchvision import datasets, transforms, utils
+from PIL import Image
+from pytorch_grad_cam import GradCAM
+from pytorch_grad_cam.utils.image import show_cam_on_image
+
+seed_everything(7)
+
+
+
+class Net_S13(nn.Module):
+#class ResNet(nn.Module):
+    def __init__(self):
+        super(Net_S13, self).__init__()
+        #super(ResNet, self).__init__()
+
+        # Control Variable
+        self.printShape = False
+
+        #Common :-
+        set1 = 64 #prepLayer
+        set2 = 128 #Layer2
+        set3 = 256 #Layer3
+        set4 = 512 #Layer4
+        avg = 1024 #channels
+        drop = 0.1 #dropout
+        S = 1 #stride
+        K = 3 #kernel_size
+
+        # PrepLayer - Conv 3x3 s1, p1) >> BN >> RELU [64k]
+        I = 3
+        O = set1
+        P = 1 #padding
+        self.prepLayer = self.convBlock(in_channels = I, out_channels = O, kernel_size = K, stride = S, padding = P)
+
+        # Layer1 -
+        # X = Conv 3x3 (s1, p1) >> MaxPool2D >> BN >> RELU [128k]
+        # R1 = ResBlock( (Conv-BN-ReLU-Conv-BN-ReLU))(X) [128k]
+        # Add(X, R1)
+        I = O
+        O = set2
+        P = 1 #padding
+        self.Layer1 = self.convMPBlock(in_channels = I, out_channels = O, kernel_size = K, stride = S, padding = P)
+
+        I = O
+        O = I
+        P = 1 #padding
+        self.resNetLayer1Part1 = self.convBlock(in_channels = I, out_channels = O, kernel_size = K, stride = S, padding = P)
+
+        I = O
+        O = I
+        P = 1 #padding
+        self.resNetLayer1Part2 = self.convBlock(in_channels = I, out_channels = O, kernel_size = K, stride = S, padding = P)
+
+        # Layer 2 -
+        # Conv 3x3 [256k]
+        # MaxPooling2D
+        # BN
+        # ReLU
+        I = O
+        O = set3
+        P = 1 #padding
+        self.Layer2 = self.convMPBlock(in_channels = I, out_channels = O, kernel_size = K, stride = S, padding = P)
+
+        # Layer 3 -
+        # X = Conv 3x3 (s1, p1) >> MaxPool2D >> BN >> RELU [512k]
+        # R2 = ResBlock( (Conv-BN-ReLU-Conv-BN-ReLU))(X) [512k]
+        # Add(X, R2)
+        I = O
+        O = set4
+        P = 1 #padding
+        self.Layer3 = self.convMPBlock(in_channels = I, out_channels = O, kernel_size = K, stride = S, padding = P)
+
+        I = O
+        O = I
+        P = 1 #padding
+        self.resNetLayer2Part1 = self.convBlock(in_channels = I, out_channels = O, kernel_size = K, stride = S, padding = P)
+
+        I = O
+        O = I
+        P = 1 #padding
+        self.resNetLayer2Part2 = self.convBlock(in_channels = I, out_channels = O, kernel_size = K, stride = S, padding = P)
+
+        # MaxPooling with Kernel Size 4
+        self.pool = nn.MaxPool2d(kernel_size = 4, stride = 4)
+
+        # FC Layer
+        I = 512
+        O = 10
+        self.lastLayer = nn.Linear(I, O)
+
+        self.aGAP = nn.AdaptiveAvgPool2d((1, 1))
+        self.flat = nn.Flatten(1, -1)
+        self.gap = nn.AvgPool2d(avg)
+        self.drop = nn.Dropout(drop)
+
+    # convolution Block
+    def convBlock(self, in_channels, out_channels, kernel_size, stride, padding, last_layer = False, bias = False):
+      if(False == last_layer):
+        return nn.Sequential(
+            nn.Conv2d(in_channels = in_channels, out_channels = out_channels, stride = stride, padding = padding, kernel_size = kernel_size, bias = bias),
+            nn.BatchNorm2d(out_channels),
+            nn.ReLU())
+      else:
+        return nn.Sequential(
+            nn.Conv2d(in_channels = in_channels, out_channels = out_channels, stride = stride, padding = padding, kernel_size = kernel_size, bias = bias))
+
+    # convolution-MP Block
+    def convMPBlock(self, in_channels, out_channels, kernel_size, stride, padding, bias = False):
+        return nn.Sequential(
+            nn.Conv2d(in_channels = in_channels, out_channels = out_channels, stride = stride, padding = padding, kernel_size = kernel_size, bias = bias),
+            nn.MaxPool2d(kernel_size = 2, stride = 2),
+            nn.BatchNorm2d(out_channels),
+            nn.ReLU())
+
+    def printf(self, n, x, string1=""):
+      if(self.printShape):
+        print(f"{n} " f"{x.shape = }" f" {string1}") ##  Comment / Uncomment this line towards the no need of print or needed print
+        pass
+    def printEmpty(self,):
+      if(self.printShape):
+        print("") ##  Comment / Uncomment this line towards the no need of print or needed print
+        pass
+
+    def forward(self, x):
+        self.printf(0.0, x, "prepLayer input")
+        x = self.prepLayer(x)
+        x = self.drop(x)
+        self.printf(0.1, x, "prepLayer output")
+        self.printEmpty()
+
+        self.printf(1.0, x, "Layer1 input")
+        x = self.Layer1(x)
+        self.printf(1.1, x, "Layer1 output --> sacroscant")
+        y = x #sacrosanct path1
+        self.printf(1.2, x, "Layer1 resnet input")
+        x = self.resNetLayer1Part1(x) #residual path1
+        x = self.drop(x)
+        x = self.resNetLayer1Part2(x) #residual path1
+        self.printf(1.3, x, "Layer1 resnet output")
+        x = x + y  #adding sacrosanct path1 and residual path1
+        x = self.drop(x)
+        self.printf(1.4, x, "res+sacrosanct output")
+        self.printEmpty()
+
+        self.printf(2.0, x, "Layer2 input")
+        x = self.Layer2(x)
+        x = self.drop(x)
+        self.printf(2.1, x, "Layer2 output")
+        self.printEmpty()
+
+        self.printf(3.0, x, "Layer3 input")
+        x = self.Layer3(x)
+        self.printf(3.1, x, "Layer3 output --> sacroscant")
+        y = x  #sacrosanct path2
+        self.printf(3.2, x, "Layer3 resnet input")
+        x = self.resNetLayer2Part1(x) #residual path2
+        x = self.drop(x)
+        x = self.resNetLayer2Part2(x) #residual path2
+        self.printf(3.3, x, "Layer3 resnet output")
+        x = x + y #adding sacrosanct path2 and residual path2
+        x = self.drop(x)
+        self.printf(3.4, x, "res+sacrosanct output")
+        self.printEmpty()
+
+        self.printf(4.0, x, "pool input")
+        x = self.pool(x)
+        self.printf(4.1, x, "pool output")
+        self.printEmpty()
+
+        # x = x.view(-1, 10)
+        self.printf(4.2, x, "For showing before last layer")
+        x = x.view(x.size(0), -1)
+        self.printf(5.0, x, "last layer input") #512, 1, 1
+        x = self.lastLayer(x)
+        # x = self.gap(x)
+        self.printf(5.1, x, "last layer output") #10, 1, 1
+        self.printEmpty()
+
+        # self.printf(7.0, x)
+        return F.log_softmax(x)
+        
+def create_model():
+    model = Net_S13()
+    return model
+    
+    
+
+class LitResnet(LightningModule):
+    def __init__(self, lr=0.05):
+        super().__init__()
+
+        self.save_hyperparameters()
+        self.model = create_model()
+
+    def forward(self, x):
+        out = self.model(x)
+        return F.log_softmax(out, dim=1)
+
+    def training_step(self, batch, batch_idx):
+        x, y = batch
+        logits = self(x)
+        loss = F.nll_loss(logits, y)
+        self.log("train_loss", loss)
+        return loss
+
+    def evaluate(self, batch, stage=None):
+        x, y = batch
+        logits = self(x)
+        loss = F.nll_loss(logits, y)
+        preds = torch.argmax(logits, dim=1)
+        acc = accuracy(preds, y, task='MULTICLASS', num_classes=10)
+
+        if stage:
+            self.log(f"{stage}_loss", loss, prog_bar=True)
+            self.log(f"{stage}_acc", acc, prog_bar=True)
+
+    def validation_step(self, batch, batch_idx):
+        self.evaluate(batch, "val")
+
+    def test_step(self, batch, batch_idx):
+        self.evaluate(batch, "test")
+
+    def configure_optimizers(self):
+        optimizer = torch.optim.SGD(
+            self.parameters(),
+            lr=self.hparams.lr,
+            momentum=0.9,
+            weight_decay=5e-4,
+        )
+        steps_per_epoch = 45000 // BATCH_SIZE
+        scheduler_dict = {
+            "scheduler": OneCycleLR(
+                optimizer,
+                0.1,
+                epochs=self.trainer.max_epochs,
+                steps_per_epoch=steps_per_epoch,
+            ),
+            "interval": "step",
+        }
+        return {"optimizer": optimizer, "lr_scheduler": scheduler_dict}

misclas_helper.py

@@ -0,0 +1,141 @@
+import os
+import math
+import numpy as np
+import pandas as pd
+import seaborn as sn
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import torchvision
+import matplotlib.pyplot as plt
+import torch.nn as nn
+import torch.nn.functional as F
+from pl_bolts.datamodules import CIFAR10DataModule
+from pl_bolts.transforms.dataset_normalizations import cifar10_normalization
+from pytorch_lightning import LightningModule, Trainer, seed_everything
+from pytorch_lightning.callbacks import LearningRateMonitor
+from pytorch_lightning.callbacks.progress import TQDMProgressBar
+from pytorch_lightning.loggers import CSVLogger
+from torch.optim.lr_scheduler import OneCycleLR
+from torch.optim.swa_utils import AveragedModel, update_bn
+from torchmetrics.functional import accuracy
+from pytorch_lightning.callbacks import ModelCheckpoint
+from torchvision import datasets, transforms, utils
+from PIL import Image
+from pytorch_grad_cam import GradCAM
+from pytorch_grad_cam.utils.image import show_cam_on_image
+
+# Denormalize the data using test mean and std deviation
+inv_normalize = transforms.Normalize(
+    mean=[-0.50/0.23, -0.50/0.23, -0.50/0.23],
+    std=[1/0.23, 1/0.23, 1/0.23]
+)
+
+
+def get_misclassified_data2(model, device, count):
+    """
+    Function to run the model on test set and return misclassified images
+    :param model: Network Architecture
+    :param device: CPU/GPU
+    :param test_loader: DataLoader for test set
+    """
+
+    PATH_DATASETS = os.environ.get("PATH_DATASETS", ".")
+    BATCH_SIZE = 256 if torch.cuda.is_available() else 64
+    NUM_WORKERS = int(os.cpu_count() / 2)
+
+
+    train_transforms = torchvision.transforms.Compose(
+        [
+            torchvision.transforms.RandomCrop(32, padding=4),
+            torchvision.transforms.RandomHorizontalFlip(),
+            torchvision.transforms.ToTensor(),
+            cifar10_normalization(),
+        ]
+    )
+
+    test_transforms = torchvision.transforms.Compose(
+        [
+            torchvision.transforms.ToTensor(),
+            cifar10_normalization(),
+        ]
+    )
+
+    cifar10_dm = CIFAR10DataModule(
+        data_dir=PATH_DATASETS,
+        batch_size=BATCH_SIZE,
+        num_workers=NUM_WORKERS,
+        train_transforms=train_transforms,
+        test_transforms=test_transforms,
+        val_transforms=test_transforms,
+    )
+
+    cifar10_dm.prepare_data()
+    cifar10_dm.setup()
+    test_loader = cifar10_dm.test_dataloader()
+
+    # Prepare the model for evaluation i.e. drop the dropout layer
+    model.eval()
+
+    # List to store misclassified Images
+    misclassified_data = []
+
+    # Reset the gradients
+    with torch.no_grad():
+        # Extract images, labels in a batch
+        for data, target in test_loader:
+
+            # Migrate the data to the device
+            data, target = data.to(device), target.to(device)
+
+            # Extract single image, label from the batch
+            for image, label in zip(data, target):
+
+                # Add batch dimension to the image
+                image = image.unsqueeze(0)
+
+                # Get the model prediction on the image
+                output = model(image)
+
+                # Convert the output from one-hot encoding to a value
+                pred = output.argmax(dim=1, keepdim=True)
+
+                # If prediction is incorrect, append the data
+                if pred != label:
+                    misclassified_data.append((image, label, pred))
+
+                if len(misclassified_data) > count :
+                  break
+    return misclassified_data
+
+
+# Yes - This is important predecessor2 for gradioMisClass
+
+def display_cifar_misclassified_data(data: list,
+                                     classes: list[str],
+                                     inv_normalize: transforms.Normalize,
+                                     number_of_samples: int = 10):
+    """
+    Function to plot images with labels
+    :param data: List[Tuple(image, label)]
+    :param classes: Name of classes in the dataset
+    :param inv_normalize: Mean and Standard deviation values of the dataset
+    :param number_of_samples: Number of images to print
+    """
+    fig = plt.figure(figsize=(10, 10))
+    img = None
+    x_count = 5
+    y_count = 1 if number_of_samples <= 5 else math.floor(number_of_samples / x_count)
+
+    for i in range(number_of_samples):
+        plt.subplot(y_count, x_count, i + 1)
+        img = data[i][0].squeeze().to('cpu')
+        img = inv_normalize(img)
+        plt.imshow(np.transpose(img, (1, 2, 0)))
+        plt.xticks([])
+        plt.yticks([])
+    plt.savefig('imshow_output_misclas.png')
+    return 'imshow_output_misclas.png'
+
+# Plot the misclassified data
+

weights_92.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e2a9fff8b371c9438f23ff81373c5858e6d5339b00ea6845c61bc6c3ef4abc0
+size 52633065