Scene Retrieval on 3RScan (test)
95MRRSGAligner
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SGAlignerModalities=Full (P + S + R + IAL), Scene Graph Input Type=Ground Truth2023.04 | 95 | 92.3 | 95.7 | 97.4 | 98.2 | 98.7 | |
| SGAlignerModalities=P + S + R, Scene Graph Input Type=Ground Truth2023.04 | 91.1 | 86.1 | 91.6 | 94.7 | 96.1 | 97 | |
| SGAlignerModalities=P + S, Scene Graph Input Type=Ground Truth2023.04 | 89.7 | 85.2 | 89.9 | 93.1 | 94.5 | 95.5 | |
| SGAlignerModalities=P, Scene Graph Input Type=Ground Truth2023.04 | 88.4 | 83.5 | 88.6 | 92.1 | 93.8 | 95.1 | |
| SGAlignerModalities=Full (P + S + R + IAL), Scene Graph Input Type=Predicted2023.04 | 88.2 | 83.3 | 88.1 | 91.8 | 93.7 | 95.1 | |
| EVAScene Graph Input Type=Ground Truth2023.04 | 86.7 | 79 | 88.4 | 93.8 | 96.3 | 97.7 | |
| VLAD + KNNSubsampling %=02023.04 | 55.7 | — | — | — | — | — | |
| SGAlignerSubsampling %=02023.04 | 38.3 | — | — | — | — | — | |
| SGAlignerSubsampling %=102023.04 | 35.6 | — | — | — | — | — | |
| SGAlignerSubsampling %=202023.04 | 34.3 | — | — | — | — | — | |
| SGAlignerSubsampling %=302023.04 | 33.9 | — | — | — | — | — | |
| VLAD + KNNSubsampling %=102023.04 | 31.6 | — | — | — | — | — | |
| SGAlignerSubsampling %=502023.04 | 31.2 | — | — | — | — | — | |
| VLAD + KNNSubsampling %=202023.04 | 27.6 | — | — | — | — | — | |
| VLAD + KNNSubsampling %=302023.04 | 22.2 | — | — | — | — | — | |
| VLAD + KNNSubsampling %=502023.04 | 16.2 | — | — | — | — | — |