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

Kadluby #46

Closed pierotofy closed 1 year ago

pierotofy commented 1 year ago

Ref #45

u4gbot commented 1 year ago
brighton_beach_small Overall accuracy from 58.73% to 63.37% | human_made_object | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 18.32% | 96.99% | 18.43% | 0.31 | | After | 15.56% | 87.60% | 15.91% | 0.27 | | Diff | -2.76% | -9.39% | -2.52% | -0.04 | | low_vegetation | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 32.94% | 33.18% | 97.89% | 0.50 | | After | 28.81% | 29.19% | 95.61% | 0.45 | | Diff | -4.13% | -3.98% | -2.28% | -0.05 | | building | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 0.90% | 49.21% | 0.91% | 0.02 | | After | 1.19% | 56.53% | 1.20% | 0.02 | | Diff | 0.29% | 7.32% | 0.30% | 0.01 | | ground | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 58.20% | 59.35% | 96.78% | 0.74 | | After | 63.07% | 64.35% | 96.94% | 0.77 | | Diff | 4.87% | 5.00% | 0.16% | 0.04 |
sheffield_park_small Overall accuracy from 80.71% to 78.92% | human_made_object | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 4.55% | 96.52% | 4.55% | 0.09 | | After | 4.17% | 95.72% | 4.17% | 0.08 | | Diff | -0.38% | -0.80% | -0.38% | -0.01 | | low_vegetation | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 82.71% | 93.75% | 87.54% | 0.91 | | After | 82.73% | 96.08% | 85.62% | 0.91 | | Diff | 0.02% | 2.33% | -1.92% | 0.00 | | building | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 9.67% | 9.68% | 98.63% | 0.18 | | After | 4.99% | 4.99% | 97.99% | 0.09 | | Diff | -4.69% | -4.69% | -0.65% | -0.08 | | ground | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 85.94% | 86.71% | 98.98% | 0.92 | | After | 83.70% | 84.34% | 99.10% | 0.91 | | Diff | -2.24% | -2.36% | 0.12% | -0.01 |

Average (2 datasets)

Overall accuracy from 69.72% to 71.14%

human_made_object Accuracy Recall Precision F1
Before 11.44% 96.76% 11.49% 0.20
After 9.86% 91.66% 10.04% 0.17
Diff -1.57% -5.10% -1.45% -0.02
low_vegetation Accuracy Recall Precision F1
Before 57.83% 63.46% 92.72% 0.70
After 55.77% 62.64% 90.62% 0.68
Diff -2.06% -0.83% -2.10% -0.02
building Accuracy Recall Precision F1
Before 5.29% 29.45% 49.77% 0.10
After 3.09% 30.76% 49.59% 0.06
Diff -2.20% 1.31% -0.18% -0.04
ground Accuracy Recall Precision F1
Before 72.07% 73.03% 97.88% 0.83
After 73.39% 74.35% 98.02% 0.84
Diff 1.31% 1.32% 0.14% 0.01