---
license: apache-2.0
library_name: pytorch
pipeline_tag: other
tags:
  - hlm-spatial
  - polynomial-hopfield
  - point-cloud-classification
  - modelnet10
  - 3d-perception
  - research-baseline
datasets:
  - modelnet10
---

# HLM-Spatial ModelNet10 Small

HLM-Spatial ModelNet10 Small is a compact polynomial-Hopfield point-cloud classifier for the ModelNet10 object classification benchmark.

This is a research baseline release, not a state-of-the-art point-cloud model. Its purpose is to make the small HLM-Spatial checkpoint inspectable and reproducible before larger follow-up spatial runs are published.

## Results

| Field | Value |
|---|---:|
| Parameters | small SpatialHLM configuration |
| Dataset | ModelNet10 |
| Task | 10-class point-cloud classification |
| Reported validation accuracy | 83.59% |
| Reported validation loss | 0.9176 |
| Checkpoint epoch | 34 |
| Encoder | pointwise |
| Points | 1024 |

## Classes

`bathtub`, `bed`, `chair`, `desk`, `dresser`, `monitor`, `night_stand`, `sofa`, `table`, `toilet`.

## Checkpoint Format

`model.pt` is a sanitized PyTorch checkpoint containing:

- `model_state`: model weights
- `config`: public architecture and task metadata
- `class_names`: ModelNet10 class names
- `val_acc`, `val_loss`, `epoch`: reported checkpoint metrics

No optimizer state, training-control state, local paths, training logs, private data, or raw run metadata are included.

## Intended Use

Use this model as a baseline for HLM-Spatial research and point-cloud classification experiments. It should not be used for safety-critical robotics, industrial inspection, medical use, or autonomous decisions.

## Limitations

- Baseline research checkpoint, not SOTA.
- Evaluated on ModelNet10-style point-cloud data.
- Does not claim real-world sensor robustness.
- Does not include a full inference package in this model repository.