Hybrid ViT-CNN Model for Forgery Detection
Budget: ₹3,000 – ₹5,000 INR
I'm looking for a machine learning expert who can develop a Hybrid ViT-CNN model to detect copy-move forgery in images that have undergone various post-processing techniques.
Key Requirements:
- Combine the Vision Transformer (ViT) with a Convolutional Neural Network (CNN) model (like ResNet) for feature extraction before passing the images to the transformer.
- Implement self-attention mechanisms in ViT to capture global contextual information.
- Detect if the copied region has gone through transformations such as scaling, rotation, and flipping using feature matching techniques.
- Identify additional post-processing transformations including blurring, color adjustments, and noise additions.
Skills & Experience:
- Proficient in machine learning, specifically in computer vision and image processing.
- Extensive experience with CNNs and Vision Transformers.
- Familiarity with feature matching techniques.
- Previous work on image forgery detection is a plus.
- Ability to work with publicly available datasets.
Please note that I will be using a publicly available dataset for training the model. I am looking for someone who can deliver high-quality work within a reasonable timeframe.
Key Requirements:
- Combine the Vision Transformer (ViT) with a Convolutional Neural Network (CNN) model (like ResNet) for feature extraction before passing the images to the transformer.
- Implement self-attention mechanisms in ViT to capture global contextual information.
- Detect if the copied region has gone through transformations such as scaling, rotation, and flipping using feature matching techniques.
- Identify additional post-processing transformations including blurring, color adjustments, and noise additions.
Skills & Experience:
- Proficient in machine learning, specifically in computer vision and image processing.
- Extensive experience with CNNs and Vision Transformers.
- Familiarity with feature matching techniques.
- Previous work on image forgery detection is a plus.
- Ability to work with publicly available datasets.
Please note that I will be using a publicly available dataset for training the model. I am looking for someone who can deliver high-quality work within a reasonable timeframe.