# 3D Object Detection on Omni3D SUN RGB-D

Source: [3D Object Detection on Omni3D SUN RGB-D](<https://www.sota2.com/research/sota/3d-object-detection-on-omni3d-sun-rgb-d>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T05:23:22.243Z

Evaluation of 3D object detection performance on the Omni3D SUN RGB-D dataset, reporting Average Precision (AP) and Average Recall (AR) metrics at IoU thresholds of 0.25 and 0.50.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | 3D Object Detection |
| Dataset (including any stated split) | Omni3D SUN RGB-D |
| Metric | AP@0.25 |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 7 of 7 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [AP@0.25](<https://www.sota2.com/research/sota/3d-object-detection-on-omni3d-sun-rgb-d.md?metric=AP%400.25>): Higher is better; unit: PERCENT.

- [AR@0.25](<https://www.sota2.com/research/sota/3d-object-detection-on-omni3d-sun-rgb-d.md?metric=AR%400.25>): Higher is better; unit: PERCENT.

- [AP@0.50](<https://www.sota2.com/research/sota/3d-object-detection-on-omni3d-sun-rgb-d.md?metric=AP%400.50>): Higher is better; unit: PERCENT.

- [AR@0.50](<https://www.sota2.com/research/sota/3d-object-detection-on-omni3d-sun-rgb-d.md?metric=AR%400.50>): Higher is better; unit: PERCENT.

## Recorded results

| Method | AP@0.25 | Result date | Report paper |
| --- | --- | --- | --- |
| CuTR (RGB-D) | 30.3 | Not recorded | 68fa760f1e209131cea6b337 |
| TR3D + FF | 29.1 | Not recorded | 68fa760f1e209131cea6b337 |
| FCAF | 27.1 | Not recorded | 68fa760f1e209131cea6b337 |
| TR3D | 27.1 | Not recorded | 68fa760f1e209131cea6b337 |
| CuTR (RGB only) | 21.5 | Not recorded | 68fa760f1e209131cea6b337 |
| Cube R-CNN (RGB only) | 18.9 | Not recorded | 68fa760f1e209131cea6b337 |
| ImVoxelNet (RGB only) | 14.4 | Not recorded | 68fa760f1e209131cea6b337 |

### CuTR (RGB-D)

AP@0.25: 30.3. Result ID: 699423343a6c3148491b516d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b337.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b337.md>)

Recorded evaluation configuration: {"Input Paradigm":"2D image-based","Input Modality":"RGB-D","Post-processing":"None","Detection limit":"100"}. This record does not assert independent reproduction.

### TR3D + FF

AP@0.25: 29.1. Result ID: 699423343a6c3148491b516a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b337.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71581e209131cea50b1e.md>)

[Implementation](<https://github.com/SamsungLabs/tr3d>)

Recorded evaluation configuration: {"Input Paradigm":"3D point-based","Post-processing":"NMS","Detection limit":"100"}. This record does not assert independent reproduction.

### FCAF

AP@0.25: 27.1. Result ID: 699423343a6c3148491b5168.

[Reporting paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b337.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c0f1e209131cea31cf2.md>)

[Implementation](<https://github.com/samsunglabs/fcaf3d>)

Recorded evaluation configuration: {"Input Paradigm":"3D point-based","Post-processing":"NMS","Detection limit":"100"}. This record does not assert independent reproduction.

### TR3D

AP@0.25: 27.1. Result ID: 699423343a6c3148491b5169.

[Reporting paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b337.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71581e209131cea50b1e.md>)

[Implementation](<https://github.com/SamsungLabs/tr3d>)

Recorded evaluation configuration: {"Input Paradigm":"3D point-based","Post-processing":"NMS","Detection limit":"100"}. This record does not assert independent reproduction.

### CuTR (RGB only)

AP@0.25: 21.5. Result ID: 699423343a6c3148491b516c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b337.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b337.md>)

Recorded evaluation configuration: {"Input Paradigm":"2D image-based","Input Modality":"RGB only","Post-processing":"None","Detection limit":"100"}. This record does not assert independent reproduction.

### Cube R-CNN (RGB only)

AP@0.25: 18.9. Result ID: 699423343a6c3148491b516b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b337.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ee21e209131cea42a09.md>)

[Implementation](<https://github.com/facebookresearch/omni3d>)

Recorded evaluation configuration: {"Input Paradigm":"2D image-based","Input Modality":"RGB only","Post-processing":"NMS","Detection limit":"100"}. This record does not assert independent reproduction.

### ImVoxelNet (RGB only)

AP@0.25: 14.4. Result ID: 699423343a6c3148491b5167.

[Reporting paper](<https://www.sota2.com/research/paper/68fa760f1e209131cea6b337.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6b101e209131cea2be2f.md>)

[Implementation](<https://github.com/saic-vul/imvoxelnet>)

Recorded evaluation configuration: {"Input Paradigm":"3D point-based","Input Modality":"RGB only","Post-processing":"NMS","Detection limit":"100"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
