centbueze/nigeria-signs-language

Computer vision project for Nigeria Sign Language gesture recognition. # Nigerian Sign Language (NSL) Recognition with YOLOv8 and ESP32-CAM This project recognizes **Nigerian Sign Language (NSL)** hand gestures using an **ESP32-CAM** for data collection and a **YOLOv8** model for training/inference. A **Flask web app** serves real-time predictions from the trained model. 🔗 **Main Repository:** github.com --- ## 📌 Project Overview - 📷 **Data Collection:** Images captured with **ESP32-CAM** (programmed via **Arduino IDE**). - 📝 **Annotation:** Labeled with LabelImg/Roboflow in **YOLO format**. - 🤖 **Model Training:** Trained a **YOLOv8** model producing `best.pt`. - 🌍 **Deployment:** **Flask** API for real-time hand gesture recognition. - 🎯 **Goal:** Practical NSL recognition to support accessibility & inclusion. --- ## 📂 Repository Structure nigeria-signs-language/ │── app.py # Flask server (inference) │── requirements.txt # Python dependencies │── README.md │── .gitignore │── dataset/ # Optional: sample or link to dataset │ ├── train/images/ train/labels/ │ └── val/images/ val/labels/ │── output/ │ └── nsl_yolo_train2/ │ └── weights/ │ └── best.pt # Trained YOLOv8 weights (provide or link) │── templates/ # Flask HTML (if UI is used) │── static/ # CSS/JS/assets (if UI is used) │── esp32_cam/ # Arduino sketches & notes (data collection) --- ## 📜 requirements.txt (minimal) flask ultralytics opencv-python numpy torch torchvision --- ## 🧠 Model Path (important) from ultralytics import YOLO model_path = r"output/nsl_yolo_train2/weights/best.pt" model = YOLO(model_path) --- ## 🧪 Quick Inference (CLI) yolo predict model=output/nsl_yolo_train2/weights/best.pt source=0 # or yolo predict model=output/nsl_yolo_train2/weights/best.pt source="sample.jpg" --- ## 🎯 Training yolo detect train data=dataset/data.yaml model=yolov8n.pt epochs=50 imgsz=640 --- ## 📄Dataset/data.yaml path: ./dataset train: train/images val: val/images nc: 26 names: [A, B, C, D, E, F, G, H, I, J, K, …