infer.py

# from model import CosineSimilarity,Discriminator
from dataloader import FlowerPairDataset
from tqdm import tqdm

import config

import torch
import pandas as pd
import torchvision.transforms as transforms
import numpy as np
from sklearn.metrics import roc_auc_score, f1_score, precision_score, recall_score, accuracy_score
from sklearn.neighbors import KNeighborsClassifier
from sklearn.svm import SVC

import torch.nn as nn
import torch.nn.functional as F
import numpy as np
from sklearn.neighbors import KNeighborsClassifier
from sklearn.metrics import roc_auc_score, f1_score


def evaluate_metrics(model, dataloader, device='cuda'):
    model.eval()
    true_labels = []
    predicted_probs = []
    cos_sim_label_1 = []
    cos_sim_label_0 = []

    with torch.no_grad():
        for x1, x2, labels in tqdm(dataloader):
            x1, x2 = x1.to(config.DEVICE), x2.to(config.DEVICE)

            # Forward pass through the model
            feature_out1, y_hat1 = model(x1)
            feature_out2, y_hat2 = model(x2)

            similarities = torch.nn.functional.cosine_similarity(feature_out1, feature_out2)
            predicted_probs.extend(similarities.cpu().numpy())
            true_labels.extend(labels.cpu().numpy())

            for i in range(len(labels)):
                if labels[i] == 1:
                    cos_sim_label_1.append(similarities[i].item())
                else:
                    cos_sim_label_0.append(similarities[i].item())

    true_labels = np.array(true_labels)
    predicted_probs = np.array(predicted_probs)
    normalized_probs = (predicted_probs + 1) / 2

    threshold = 0.4
    predicted_labels = (predicted_probs > threshold).astype(int)
    roc_auc = roc_auc_score(true_labels, normalized_probs)
    f1 = f1_score(true_labels, predicted_labels)
    precision = precision_score(true_labels, predicted_labels)
    recall = recall_score(true_labels, predicted_labels)
    accuracy = accuracy_score(true_labels,predicted_labels)

    mean_cos_sim_label_1 = np.mean(cos_sim_label_1)
    mean_cos_sim_label_0 = np.mean(cos_sim_label_0)

    return roc_auc , f1 , precision , recall , accuracy , mean_cos_sim_label_1 , mean_cos_sim_label_0

# # import utils
# transform = transforms.Compose([
#     transforms.Resize((128, 128)),
#     transforms.ToTensor(),
#     transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
# ])
# model = Dis(num_classes=82,num_features=2048)
# df = pd.read_csv('Dataset/test_pairs.csv')
# dataset = FlowerPairDataset(df=df,transform=transform)
# loader = torch.utils.data.DataLoader(dataset,batch_size=32,shuffle=False)
# # model = Discriminator(num_classes=82,num_features=2048).to(config.DEVICE)
# # utils.load_checkpoint(checkpoint_path='logs/ResnetWithArcCenter.pt',model=model)
# print(evaluate_metrics(model=model,dataloader=loader))