Feat/Backend: Add Depth Estimation for Object Proximity Warnings-147

[tamannaa-rath]

🔗 Related Issue

Closes #147

---

📝 Summary of Changes

This PR adds a new DepthEstimator module to handle smart distance checking for industrial hazards.

• What it does: It takes a video frame and a bounding box from an object detector, passes it through a lightweight deep learning model (MiDaS_small), and calculates the average relative distance of that specific object.
• Why it matters: It prevents injuries by triggering a system alert if a dangerous object (like a knife or flame) gets too close to the camera.
• Decoupled Architecture: Because the main video pipeline code is still being written by another contributor (who has raised a PR), this module is built independently. It uses a strict input-output agreement so it will seamlessly connect to their work as soon as their PR merges.

---

🔍 Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)
• ✨ New feature (non-breaking change that adds functionality)
• 💥 Breaking change (fix or feature that would cause existing functionality to not work as expected)
• 📖 Documentation update
• 🧪 Test addition or improvement
• 🧹 Refactor / Code cleanup (no functional change)
• ⚙️ Build / CI / Chore

---

🧪 How Was This Tested?

I ran the automated unit testing suite locally to check the initialization logic, min-max image math normalization, matrix array slicing coordinates, and the proximity alert limits.

python -m unittest core/physical/test_depth_estimator.py

Test environment:

• OS: Windows 11 (Ryzen 7, Integrated AMD Graphics)
• Python version: 3.12+
• Node version (if frontend): N/A

---

---

✅ Pre-Submission Checklist

• My branch is up to date with upstream/main
• My code follows the project's coding standards (Black, isort, flake8 pass)
• I have written / updated tests for my changes
• All existing tests pass (python -m pytest tests/)
• I have added docstrings and type hints to all new public functions
• I have updated relevant documentation (README, docs/, inline comments)
• My PR title follows the Conventional Commits format
• I have not committed any .env files, secrets, or model weights
• This PR addresses only the scope of the linked issue

---

💬 Additional Notes for Reviewer

• Hardware-Adaptive Logic: The code automatically detects the user's computer specs. It safely falls back to CPU processing if an NVIDIA GPU isn't available, meaning it can be run locally without system crashes.

• No-Download Testing: The test suite uses unittest.mock to completely bypass downloading the heavy 300MB model files from the internet during test execution.