Assembly Instruction Generation

This repository contains the implementation of Furniture Assembly Instruction Generation using a fine-tuned T5 Transformer model. The system takes a structured Bill of Materials (BoM) for a furniture item and generates coherent, step-by-step natural language assembly instructions.

---

Repository Structure

• 100k_Training_Furniture_Dataset.zip – Training dataset (100,000 furniture examples)
• 10K_Furniture_Testing_Data.zip – Testing dataset (10,000 furniture examples)
• Furniture100K_Model_Train_Test.ipynb – Main training and testing notebook
• Furniture_Assembly_Model_Training.ipynb – Additional model training notebook
• TRAINING MODEL/ – Contains a Google Drive link to the trained model
• additionNN.ipynb – Addition neural network implementation

---

How to Use

1. Open Furniture100K_Model_Train_Test.ipynb in Jupyter Notebook or Google Colab.

Update the dataset paths in Furniture100K_Model_Train_Test.ipynb to point to the extracted folders:

train_data_path = "100k_Training_Furniture_Dataset/"
test_data_path = "10K_Furniture_Testing_Data/"

3. Run all cells to train and/or test the model.

---

Model Overview

• Architecture: T5-small, fine-tuned for sequence-to-sequence instruction generation
• Training Data: 100,000 BoM–instruction pairs
• Evaluation Metrics:
• BLEU: 80.38
• ROUGE-L: 0.927

---

Example

Input (BoM): (Bedside_table (Tabletop) (Legs) (Drawers) (Fasteners))

Generated Output: Combine Tabletop + Legs -> Tabletop_Legs_assembly Combine Tabletop_Legs_assembly + Drawers -> Tabletop_Legs_assembly_Drawers_assembly Combine Tabletop_Legs_assembly_Drawers_assembly + Fasteners -> Finished_Tabletop_Legs_assembly_Drawers_assembly_Fasteners_assembly

---

Key Features

• Automatically converts structured BoMs into detailed, step-by-step assembly instructions
• Large-scale dataset with 100K (BoM, instruction) pairs
• High BLEU and ROUGE scores, indicating strong performance
• Flexible and extendable to other instruction-generation domains