Projects
Research Projects
Deepfake Detection using CNN
Goal: Develop a CNN-based method trained on ensemble data to detect manipulated videos/images.
ANT-AgroAI
Goal: Use real-world Nigerian farm data to classify plant and animal diseases using deep learning.
Methods: CNN, multi-class labelling, mobile deployment, custom dataset
Datasets
ANT-AkwaSpeech v1.0
Speech dataset for Ibibio/English accents in Southern Nigeria.
License: Dataset License
Download / Request AccessANT-SymptomText v1.0
Text dataset of symptom descriptions in Pidgin English, Ibibio.
- Annotated medical keywords
- Transliteration and code-switching examples
License: Dataset License
Download / Request AccessOpen-source Tools
Sio-Iso
Python script to ensure anonymity using face detection + Gaussian blur.
- OpenCV + Haar Cascades
- Accepts .mp4 input, Returns .avi output
- Logs faces detected
Nsin-Uyom
Python script to inject realistic audio/visual noise into training data for realistic scenarios.
- Simulates in background chatter into clean sound profiles
- Supports WAV and MP3 input formats
- Noise profiles from real-world conditions
Under Development Projects
ANT-AgroAI
Detecting crop & livestock diseases using real photos captured across rural farms in Nigeria.
- Multi-label CNN classifier
- Mobile-first design
Rural Biometric Verification
Low-bandwidth, privacy-conscious biometric ID system for rural Nigerian communities.
- Edge-computing support
- Designed for poor lighting + connectivity