Add README for Robust6DNet

[Copilot]

No documentation existed for the project. Adds a structured README.md covering the full pipeline from dataset setup through training to inference.

Contents

• Overview — Three-stage pipeline: KeypointNet heatmap detection → 2D–3D correspondence → PnP pose estimation
• Repository structure — Annotated directory tree
• Data generation pipeline — Ordered steps for find_high_visib_images, generate_keypoints, generate_annotations, generate_crops, generate_heatmaps, plus the gen.sh convenience script
• Training CLI — Full argument table with defaults for train_r6dnet.py
• Model architecture — EfficientNet/ResNet backbone + transposed-conv upsampling head + per-channel min-max normalization
• Evaluation metrics — MSE, PCK, MDE, ADD, ADD-S, reprojection error
• Pre-trained models — Table of included weights for objects 1, 2, 8, 11
• References — Linemod (ACCV 2012) and BOP benchmark (ECCV 2018)

Original prompt

Create a detailed and structured but straightforward README about this project

---

🔒 GitHub Advanced Security automatically protects Copilot coding agent pull requests. You can protect all pull requests by enabling Advanced Security for your repositories. Learn more about Advanced Security.