The Problem
The client required an accurate system to recognize images of specific individuals from a known set of faces (Ajay Bidyarthy, Udit Narayan, and Aditya Narayan). The system needed to distinguish between these individuals and unknown faces.
Our Solution
We implemented a facial recognition system using DeepFace with the pre-trained VGG-Face model. The system processes input images, compares them with a database of known images, and identifies the person if they are a known match.
Solution Architecture
The architecture is built on Python with the DeepFace library. It consists of:
- Input & Image Processing: The system receives an image in JPG format.
- Feature Extraction: Using VGG-Face, facial features are extracted.
- Database Search: These features are compared to the known images of Ajay Sir, Udit Narayan, and Aditya Narayan stored in a local folder.
- Face Matching: If a match is found, the person is identified; otherwise, the face is marked as unknown.
Deliverables
- A face recognition script using DeepFace.
- Known and unknown face testing setup.
- A solution that recognizes faces from test images.
Tech Stack
- Tools used
- Â DeepFace
- OpenCV
- Matplotlib
- Language/techniques used
- Python
- Models used
- Â DeepFace VGG-Face model for face recognition
- Skills used
- Python Programming
- Machine learning
- Face recognition
- Databases used
- Â Local directory for storing known Faces & Testing Images
What are the technical Challenges Faced during Project Execution
- Incorrect Image Recognition: Initially, the system was not able to distinguish between closely related faces (e.g., Udit Narayan’s Young Age face and Aditya Narayan).
- Empty DataFrames: Some images resulted in empty DataFrames, causing index errors.
How the Technical Challenges were Solved
- Error Handling: Added checks for empty DataFrames to prevent errors when no face is detected.
- Trained model on more distinguishable images
Business Impact
The system provided the client with a reliable way to recognize specific individuals. This improved automation in their workflow and helped categorize images with high accuracy.
Project Snapshots
