Gear Vision is a deep learning model for detecting and classifying various car parts, designed to assist firms in automating and optimizing parts identification and inventory management.
- Deep Learning Framework:
- TensorFlow/Keras: Used for building and training the deep learning model with MobileNetV2 architecture.
- Image Processing:
- OpenCV: Applied for various image processing tasks.
- PIL (Pillow): Utilized for loading and resizing images.
- Data Augmentation:
- ImageDataGenerator: Employed for augmenting the dataset with transformations such as rotation, shifting, shearing, and flipping.
- Programming Language:
- Python: The primary language for data processing, model training, and evaluation.
- Additional Libraries:
- NumPy and Pandas: Used for data manipulation.
- Matplotlib: For visualizing learning curves and predictions.
The model was trained with good clear images and when tested with noisy images, it provided accurate predictions for real-time images, making it suitable for applications where customers submit live images of car parts.
 |
 |
|---|---|
| Radiator Hose | Brake Rotor |
 |
 |
|---|---|
| Engine Block | Water Pump |
- User Interface: Create a web app for customers to upload images and get instant results.
- Enhanced Processing: Implement techniques to handle various image qualities and lighting conditions.
- RAG Application: Build a RAG system to provide contextual information and FAQs alongside image classification.
These improvements will enhance customer interaction and streamline part identification and support.
This project is licensed under the MIT License.