ITEP-84336: Fix incorrect camera pose for models generated with VGGT

[daddo-intel]

📝 Description

VGGT outputs camera poses and depth in scale-ambiguous units, which cased:

1. Incorrect mesh dimenstions
2. incorrect pixels_per_meter in renderTopView
3. Inconsistent scale compared to MapAnything

This PR implements automatic metric scale for VGGT using known camera poses provided to` _processOutputs() and applies scale factor to:

1. Camera translations
2. world_points_from_depth
3. world_points

✨ Type of Change

Select the type of change your PR introduces:

• 🐞 Bug fix – Non-breaking change which fixes an issue
• 🚀 New feature – Non-breaking change which adds functionality
• 🔨 Refactor – Non-breaking change which refactors the code base
• 💥 Breaking change – Changes that break existing functionality
• 📚 Documentation update
• 🔒 Security update
• 🧪 Tests
• 🚂 CI

🧪 Testing Scenarios

Describe how the changes were tested and how reviewers can test them too:

• ✅ Tested manually
• 🤖 Ran automated end-to-end tests

✅ Checklist

Before submitting the PR, ensure the following:

• 🔍 PR title is clear and descriptive
• 📝 For internal contributors: If applicable, include the JIRA ticket number (e.g., ITEP-123456) in the PR title. Do not include full URLs
• 💬 I have commented my code, especially in hard-to-understand areas
• 📄 I have made corresponding changes to the documentation
• ✅ I have added tests that prove my fix is effective or my feature works

[Copilot]

Pull request overview

Implements automatic metric scale correction for VGGT reconstructions by propagating known camera poses/locations from the Manager → Mapping API → VGGT model, then scaling predicted camera translations and reconstructed world points to match real-world units.

Changes:

• Extend the mapping request pipeline to include per-image camera_location (pose) metadata from Manager to the mapping service.
• Compute a metric scale factor inside VGGTModel._processOutputs() and apply it to camera translations and world points (and depth).
• Adjust the example mapping client’s timeouts/polling behavior for longer-running reconstructions.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 4 comments.

File	Description
mapping/tools/client_example.py	Increases request timeout and simplifies polling behavior for async reconstructions.
mapping/src/vggt_model.py	Adds metric scaling logic for VGGT outputs and updates preprocessing/intrinsics/camera pose packaging.
mapping/src/api_service_base.py	Parses optional camera_locations form data and attaches it to per-image inference payloads.
manager/src/django/mesh_generator.py	Extracts camera pose via CamSerializer and uploads it alongside images to the mapping service.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 4 comments.

[Copilot]

The metric scale factor is computed using baseline_m (median over all provided camera_location translations) but b_units is computed only from the first two predicted camera poses. If those two frames are not representative of the median baseline (or if ordering differs), this can produce an incorrect scale factor and reintroduce the same “wrong scale” behavior.

Consider computing the predicted baseline in model units using the same robust statistic as baseline_m (e.g., median of pairwise camera-center distances across all valid pose pairs, aligned by index), then set s = median_baseline_m / median_baseline_units (or compute a median of per-pair ratios).