README.md

---
datasets:
- tbhugging/gummybear-tomography
library_name: pytorch
license: apache-2.0
metrics:
- rmse
pipeline_tag: image-feature-extraction
tags:
- tomography
- particle-localization
- cnn
- pooled
- gap
- multiview
- geometry-fusion
model-index:
- name: tbhugging/camera_orbit_compact_09_2b
  results:
  - task:
      type: image-feature-extraction
      name: Multi-view camera-orbit particle localisation (xyz)
    dataset:
      name: GummyBear Tomography (M8 catalog, M9 multi-view)
      type: tbhugging/gummybear-tomography
      config: m8_1
      split: validation
    metrics:
    - type: rmse
      value: 1.320252
      name: RMSE_total (Euclidean xyz)
    source:
      url: https://github.com/tbgitoo/gummybear-tomography/blob/master/GummyBearTomography_Final_Report.ipynb
      name: Final Report M9 Step 2 / 09_2B (pooled GAP)
  - task:
      type: image-feature-extraction
      name: Multi-view camera-orbit particle localisation (xyz)
    dataset:
      name: GummyBear Tomography (M8 catalog, M9 multi-view)
      type: tbhugging/gummybear-tomography
      config: m8_1
      split: test
    metrics:
    - type: rmse
      value: 0.860746
      name: RMSE_total (Euclidean xyz)
    source:
      url: https://github.com/tbgitoo/gummybear-tomography/blob/master/GummyBearTomography_Final_Report.ipynb
      name: Final Report M9 Step 2 / 09_2B (pooled GAP)
---

# `tbhugging/camera_orbit_compact_09_2b`

## What this model does

Given **6** camera-orbit views (0°, 60°, 120°, 180°, 240°, 300°) of 128×128 projections from the GummyBear Tomography dataset, the model predicts

- particle_x
- particle_y
- particle_z

coordinates of the embedded particle.



## Architecture

Per-view **GAP (pooled) CNN trunk** → camera **sin/cos geometry tokens** →
**compact fusion MLP** (`e2e_pooled_geometry_fusion`).

Final Report **M9 Step 2 / 09_2B** — compact head (`f2_state`).

- Hub: [tbhugging/camera_orbit_compact_09_2b](https://huggingface.co/tbhugging/camera_orbit_compact_09_2b)
- Companion dataset: [tbhugging/gummybear-tomography](https://huggingface.co/datasets/tbhugging/gummybear-tomography)
- Runnable report: [GummyBearTomography_Final_Report.ipynb](https://github.com/tbgitoo/gummybear-tomography/blob/master/GummyBearTomography_Final_Report.ipynb) (M9 Step 2 — 09_2B pooled ladder)
- Source study checkpoint: [checkpoints/m9/m09_e2e_pooled_geometry_fusion.pt](https://huggingface.co/datasets/tbhugging/gummybear-tomography/blob/main/checkpoints/m9/m09_e2e_pooled_geometry_fusion.pt) (`f2_state` / `m09_2_e2e_pooled_geometry_fusion`)

## Training configuration

- Protocol: `09_2B` (pooled GAP; excludes 09_2A Fourier)
- Backbone: `pooled_gap` (`pooled` trunk)
- Input field: `anomaly_ref`
- Normalisation: `per_image_zscore`
- Camera orbit: `0°, 60°, 120°, 180°, 240°, 300°` (6 views)
- Geometry: `sin_theta, cos_theta` (concat)
- Fusion: hidden `128`, depth `1`
- Targets: `particle_x, particle_y, particle_z`
- Builder: `GeometryAwareFourierFusionLocalizer.for_09_2_pooled()`
- Variant: `m09_2_e2e_pooled_geometry_fusion`
- Trainable parameters: `132358`
- Stage-A learning rate: `0.001`

## Evaluation Results

Structured scores for the Hub widget are declared in the YAML `model-index` / `metrics` metadata ([Model Cards — Evaluation Results](https://huggingface.co/docs/hub/model-cards#evaluation-results)).

### Testing Data

- Dataset: [tbhugging/gummybear-tomography](https://huggingface.co/datasets/tbhugging/gummybear-tomography) (config `m8_1`)
- Splits: `validation`, `test`
- Protocol: Final Report **M9 Step 2 / 09_2B** (camera orbit 0°, 60°, 120°, 180°, 240°, 300°, `anomaly_ref`, `per_image_zscore`, **GAP pooled trunk only**, compact geometry fusion; excludes 09_2A Fourier)
- Source checkpoint: [checkpoints/m9/m09_e2e_pooled_geometry_fusion.pt](https://huggingface.co/datasets/tbhugging/gummybear-tomography/blob/main/checkpoints/m9/m09_e2e_pooled_geometry_fusion.pt) (`f2_state` only)

### Metrics

Reported error is **Euclidean RMSE** over particle `(x,y,z)`:
`d_i = ||pred_i - y_i||_2`, then `RMSE_total = sqrt(mean_i d_i^2)`.
Hub metric id: `rmse` (display name `RMSE_total (Euclidean xyz)`).

### Results

Scores match the Final Report M9 Step 2 **09_2B pooled GAP** bar (not 09_2A Fourier; not element-wise MSE).

| Split | Metric | Value |
| --- | --- | ---: |
| `validation` | RMSE_total (Euclidean xyz) | `1.320252` |
| `test` | RMSE_total (Euclidean xyz) | `0.860746` |

Source: [Final Report M9 Step 2 / 09_2B](https://github.com/tbgitoo/gummybear-tomography/blob/master/GummyBearTomography_Final_Report.ipynb).

## Load

```python
import torch
# libraries from https://github.com/tbgitoo/gummybear-tomography
# Historical class name — use .for_09_2_pooled() for this GAP checkpoint only.
from tomography_ml.localization.localize_multiview import (
    GeometryAwareFourierFusionLocalizer,
)

n_views = 6
view_angles_deg = [0.0, 60.0, 120.0, 180.0, 240.0, 300.0]
model = GeometryAwareFourierFusionLocalizer.for_09_2_pooled(
    n_views=n_views,
    view_angles_deg=view_angles_deg,
)
views = torch.zeros(1, n_views, 1, 128, 128)
model(views)  # materialise lazy layers
state = torch.load('pytorch_model.bin', map_location='cpu', weights_only=True)
model.load_state_dict(state)
model.eval()
xyz = model(views)
```

Input tensor shape: `[batch, n_views, channels, height, width]`.
Do **not** load with `.for_09_2()` (that builds the 09_2A Fourier trunk).
Also excludes single-view M8, 09_3 large fusion, and M10 illumination stacks.

## Inference

For an example with worked download, model instanciation and inference, see: [11_2_test_camera_orbit_compact_09_2b.ipynb](https://github.com/tbgitoo/gummybear-tomography/blob/master/notebooks/milestone_11/11_2_test_camera_orbit_compact_09_2b.ipynb) in the [github.com/tbgitoo/gummybear-tomography](https://github.com/tbgitoo/gummybear-tomography) repository.