This repository contains the machine learning model developed for the Amazon ML Hackathon. The project utilizes YOLO (You Only Look Once) and Roboflow to perform dimensional data extraction from large-scale image datasets. The primary goal was to achieve high precision in automatically identifying and extracting size dimensions from images for various applications.
- Dimensional Data Extraction: Uses YOLO for accurate and efficient detection of object boundaries and dimensions.
- Large-Scale Image Processing: Trained and tested on over 100,000 images, demonstrating the model's scalability and robustness.
- High Accuracy: Achieved a 50% F1 score, indicating a balanced precision and recall, suitable for practical applications.
The model showed significant improvements in accuracy and a systematic reduction in loss metrics throughout the training phase:
- Training Loss: Consistently decreased, reflecting the model's increasing accuracy over time.
- Validation Accuracy: Showcased the model's effectiveness on unseen data.
- YOLO: For object detection and dimensional analysis.
- Roboflow: For image annotation and pre-processing to improve model training efficiency.
- Python: Primary programming language.
- EasyOCR: Integrated for text recognition from detected objects.
