Skip to content

README.md

Latest commit

 

History

History
48 lines (33 loc) · 1.97 KB

README.md

File metadata and controls

48 lines (33 loc) · 1.97 KB
Raw

Material Stream Identification System

An automated system for identifying material streams using feature extraction, data augmentation, and machine learning classifiers (SVM and k-NN).

🚀 Project Pipeline

The project follows a structured machine learning workflow, divided into a core data preparation phase followed by algorithm-specific implementations.

🛠 Phase 1: Data Preparation

  1. Load & Clean: Initial data ingestion and removal of noise or corrupted files.
  2. Preprocessing: Data normalization and formatting for model readiness.
  3. Train-Test Split: Partitioning the dataset into training and evaluation sets.
  4. Data Augmentation: Expanding the dataset to improve model generalization.

🧠 Model Workflows

🔹 k-Nearest Neighbors (k-NN)

Workflow designed for distance-based classification:

  • Step 1: feature_extraction_KNN — Extracting relevant spatial or color features.
  • Step 2: scaleData_KNN — Standardizing data to ensure equal feature weighting.
  • Step 3: knn_train — Training the k-NN classifier.
  • Step 4: camera_knn — Real-time identification via live camera feed.

🔸 Support Vector Machine (SVM)

Workflow optimized for high-dimensional boundary classification:

  • Step 1: feature_extraction_SVM — Extracting features tailored for hyperplane separation.
  • Step 2: scale_data_SVM — Feature scaling for optimal SVM convergence.
  • Step 3: svm_train — Training the SVM model.
  • Step 4: camera_svm — Real-time identification via live camera feed.

📊 Summary Table

Stage k-NN Path SVM Path
Features feature_extraction_KNN feature_extraction_SVM
Scaling scaleData_KNN scale_data_SVM
Training knn_train svm_train
Inference camera_knn camera_svm

Note: The preprocessing and augmentation steps are shared across both models to ensure a fair comparison of performance.