Performance Analysis of LOCALIZE IMAGE FORGERY USING END-TO-END ATTENTION NETWORK
Implementation plan:
===================
Scenario 1:(Using Casia-V2 Dataset)
==============================
Step 1: Initially, we collect and load Casia V2 Dataset.
Step 2: Then, we pre-process the collected data from Dataset.
Step 3: Next, we increase network generalization using channel attention mechanisms for both spatial and frequency domains. ( channel attention HRNet and a channel attention DCT net to extract spatial and frequency domain multi-resolution features)
Step 4: Next, we implement the Multi-Resolution Feature Extraction modern forgery detection(MFD) algorithm.
Step 5: Next, we Concatenate the extracted features from both subnetworks and predict forged regions.
step 6:Finally, we plot performance for the following metrics:
6.1: Number of epochs vs. Accuracy (%)
6.2: Number of epochs vs. Recall (%)
6.3: Number of epochs vs. Precision (%)
6.4: Number of epochs vs. F1-score (%)
6.5. Training and Validation Loss(graph)
6.5: AUC Curve
6.6: ROC Curve
6.7 IOU(Intersection over Union)-(%)
Scenario 2:(Using Columbia Dataset)
===============================
Step 1: Initially, we collect and load Columbia Dataset.
Step 2: Then, we pre-process the collected data from Dataset.
Step 3: Next, we increase network generalization using channel attention mechanism for both spatial and frequency domains.( channel attention HRNet and a channel attention DCT net to extract spatial and frequency domain multi-resolution features)
Step 4: Next, we implement the Multi-Resolution Feature Extraction modern forgery detection(MFD) algorithm.
Step 5: Next, we Concatenate the extracted features from both subnetworks and predict forged regions.
step 6:Finally, we plot performance for the following metrics:
6.1: Number of epochs vs. accuracy (%)
6.2: Number of epochs vs. Recall (%)
6.3: Number of epochs vs. Precision (%)
6.4: Number of epochs vs. F1-score (%)
6.5. Training and validation loss(graph)
6.5: AUC Curve
6.6: ROC Curve
6.7 IOU(Intersection over Union)-(%)
Scenario 3:(Using Carvalho Dataset)
==============================
Step 1: Initially, we collect and load the Carvalho Dataset.
Step 2: Then, we pre-process the collected data from Dataset.
Step 3: Next, we increase network generalization using channel attention mechanism for both spatial and frequency domains.( channel attention HRNet and a channel attention DCT net to extract spatial and frequency domain multi-resolution features)
Step 4: Next, we implement the Multi-Resolution Feature Extraction modern forgery detection(MFD) algorithm.
Step 5: Next, we Concatenate the extracted features from both subnetworks and predict forged regions.
step 6:Finally, we plot performance for the following metrics:
6.1: Number of epochs vs. accuracy (%)
6.2: Number of epochs vs. Recall (%)
6.3: Number of epochs vs. Precision (%)
6.4: Number of epochs vs. F1-score (%)
6.5. Training and validation loss(graph)
6.5: AUC Curve
6.6: ROC Curve
6.7 IOU(Intersection over Union)-(%)
Scenario 4:(Using comofod Dataset)
==============================
Step 1: Initially, we collect and load the comofod Dataset.
Step 2: Then, we pre-process the collected data from Dataset.
Step 3: Next, we increase network generalization using channel attention mechanism for both spatial and frequency domains.( channel attention HRNet and a channel attention DCT net to extract spatial and frequency domain multi-resolution features)
Step 4: Next, we implement the Multi-Resolution Feature Extraction modern forgery detection(MFD) algorithm.
Step 5: Next, we Concatenate the extracted features from both subnetworks and predict forged regions.
step 6:Finally, we plot performance for the following metrics:
6.1: Number of epochs vs. accuracy (%)
6.2: Number of epochs vs. Recall (%)
6.3: Number of epochs vs. Precision (%)
6.4: Number of epochs vs. F1-score (%)
6.5. Training and validation loss(graph)
6.5: AUC Curve
6.6: ROC Curve
6.7 IOU(Intersection over Union)-(%)
Scenario 5:(Using MICC F220 Dataset)
================================
Step 1: Initially, we collect and load the MICC F220 Dataset.
Step 2: Then, we pre-process the collected data from Dataset.
Step 3: Next, we increase network generalization using channel attention mechanism for both spatial and frequency domains.( channel attention HRNet and a channel attention DCT net to extract spatial and frequency domain multi-resolution features)
Step 4: Next, we implement the Multi-Resolution Feature Extraction modern forgery detection(MFD) algorithm.
Step 5: Next, we Concatenate the extracted features from both subnetworks and predict forged regions.
step 6:Finally, we plot performance for the following metrics:
6.1: Number of epochs vs. accuracy (%)
6.2: Number of epochs vs. Recall (%)
6.3: Number of epochs vs. Precision (%)
6.4: Number of epochs vs. F1-score (%)
6.5. Training and validation loss(graph)
6.5: AUC Curve
6.6: ROC Curve
6.7 IOU(Intersection over Union)-(%)
Scenario 6:(Using CG-1050Dataset)
=============================
Step 1: Initially, we collect and load the CG-1050 Dataset.
Step 2: Then, we pre-process the collected data from Dataset.
Step 3: Next, we increase network generalization using channel attention mechanism for both spatial and frequency domains.( channel attention HRNet and a channel attention DCT net to extract spatial and frequency domain multi-resolution features)
Step 4: Next, we implement the Multi-Resolution Feature Extraction modern forgery detection(MFD) algorithm.
Step 5: Next, we Concatenate the extracted features from both subnetworks and predict forged regions.
step 6:Finally, we plot performance for the following metrics:
6.1: Number of epochs vs. accuracy (%)
6.2: Number of epochs vs. Recall (%)
6.3: Number of epochs vs. Precision (%)
6.4: Number of epochs vs. F1-score (%)
6.5. Training and validation loss(graph)
6.5: AUC Curve
6.6: ROC Curve
6.7 IOU(Intersection over Union)-(%)
Dataset Link:
——————-
1. Casia V2 Dataset: https://www.kaggle.com/datasets/divg07/casia-20-image-tampering-detection-dataset
2. Columbia Dataset: https://www.kaggle.com/datasets/jessicali9530/coil100
3. Carvalho Dataset:https://www.kaggle.com/darrencarvalho7/code
4. CoMoFoD: https://www.kaggle.com/datasets/tusharchauhan1898/comofod
5. MICC F220: https://www.kaggle.com/datasets/mashraffarouk/micc-f220
6. CG-1050: https://www.kaggle.com/datasets/saurabhshahane/cg1050
Software Requirement:
——————————–
1. Development Tool: Python – 3.11.4 or above
2. Operating System: Windows 10 (64-bit)
Note:
——–
1) If the above plan does not satisfy your requirement, please provide the processing details, like the above step-by-step.
2) Please note that this implementation plan does not include any further steps after it is put into implementation.
3) If the above plan satisfies your requirement please confirm with us.
4) We develop simulation based projects only, not in real time
