OpenDroneMap / ODMSemantic3D

An open photogrammetry dataset of classified 3D point clouds for automated semantic segmentation. CC BY-SA 4.0
Other
15 stars 7 forks source link

Added waterbury point cloud #44

Closed HeDo88TH closed 1 year ago

HeDo88TH commented 1 year ago

As title says!

u4gbot commented 1 year ago
sheffield_park_small Overall accuracy from 79.35% to 80.71% | human_made_object | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 4.02% | 93.17% | 4.03% | 0.08 | | After | 4.55% | 96.52% | 4.55% | 0.09 | | Diff | 0.53% | 3.35% | 0.52% | 0.01 | | low_vegetation | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 80.33% | 91.98% | 86.38% | 0.89 | | After | 82.71% | 93.75% | 87.54% | 0.91 | | Diff | 2.38% | 1.77% | 1.16% | 0.01 | | ground | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 85.57% | 85.83% | 99.65% | 0.92 | | After | 85.94% | 86.71% | 98.98% | 0.92 | | Diff | 0.37% | 0.87% | -0.66% | 0.00 | | building | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 5.43% | 5.45% | 95.69% | 0.10 | | After | 9.67% | 9.68% | 98.63% | 0.18 | | Diff | 4.24% | 4.24% | 2.94% | 0.07 |
brighton_beach_small Overall accuracy from 74.45% to 58.73% | human_made_object | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 22.56% | 82.01% | 23.73% | 0.37 | | After | 18.32% | 96.99% | 18.43% | 0.31 | | Diff | -4.23% | 14.98% | -5.30% | -0.06 | | low_vegetation | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 69.99% | 73.71% | 93.26% | 0.82 | | After | 32.94% | 33.18% | 97.89% | 0.50 | | Diff | -37.04% | -40.53% | 4.63% | -0.33 | | ground | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 74.27% | 74.51% | 99.57% | 0.85 | | After | 58.20% | 59.35% | 96.78% | 0.74 | | Diff | -16.07% | -15.16% | -2.79% | -0.12 | | building | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 1.75% | 61.60% | 1.76% | 0.03 | | After | 0.90% | 49.21% | 0.91% | 0.02 | | Diff | -0.85% | -12.39% | -0.86% | -0.02 |

Average (2 datasets)

Overall accuracy from 76.90% to 69.72%

human_made_object Accuracy Recall Precision F1
Before 13.29% 87.59% 13.88% 0.22
After 11.44% 96.76% 11.49% 0.20
Diff -1.85% 9.17% -2.39% -0.02
low_vegetation Accuracy Recall Precision F1
Before 75.16% 82.85% 89.82% 0.86
After 57.83% 63.46% 92.72% 0.70
Diff -17.33% -19.38% 2.90% -0.16
ground Accuracy Recall Precision F1
Before 79.92% 80.17% 99.61% 0.89
After 72.07% 73.03% 97.88% 0.83
Diff -7.85% -7.15% -1.73% -0.06
building Accuracy Recall Precision F1
Before 3.59% 33.53% 48.73% 0.07
After 5.29% 29.45% 49.77% 0.10
Diff 1.70% -4.08% 1.04% 0.03