Capstone Project - Two-Dimensional Convolutional Neural Network (2D CNN) for Seam Carving Detection

Description 📋

Full article on LinkedIn.

Seam Carving (or liquid rescaling) is an algorithm for content-aware image resizing, developed by Shai Avidan and Ariel Shamir.

Capstone project was based on the construction of a 2D CNN together with image processing techniques with the objective of detecting image tampering caused by the Seam Carving technique, taught by PhD Kelton Augusto Pontara da Costa from São Paulo State University (Unesp).

---

Content: "Section 5 - Deep Learning aimed at Image Tamper Detection" 🛠️

Full article on LinkedIn.

Source code referring to "Section 5 - Deep Learning aimed at Image Tamper Detection" of the article referenced above:

1. Data Collection and Pre-processing;
2. Modeling, Training, Optimization, Evaluation - Deep Learning Technique;
3. Communication and Results.

---

Techniques and Technologies Used 🖥️

• ``Supervised Learning``
• ``Multiclass Classification``
• ``Two-Dimensional Convolutional Neural Network (2D CNN)``
• ``Local Binary Pattern (LBP)``
• ``Seam Carving/Seam Insertion``
• ``Python (os, itertools, cv2, matplotlib, numpy, pandas, tensorflow, keras, sklearn, skimage, scipy, image, PIL, seam_carving)``
• ``Sam Houston State University (SHSU) Database``
• ``Google Colab``

---

Developer 🧑‍💻

Gabriel Ferreira