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 Wietrznia point cloud #39

Closed HeDo88TH closed 1 year ago

HeDo88TH commented 1 year ago

Using sampled point cloud with 0.10m

u4gbot commented 1 year ago
brighton_beach_small Overall accuracy from 92.26% to 63.73% | ground | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 92.45% | 93.34% | 98.98% | 0.96 | | After | 63.73% | 64.12% | 99.06% | 0.78 | | Diff | -28.72% | -29.22% | 0.07% | -0.18 | | building | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 10.34% | 99.56% | 10.34% | 0.19 | | After | 1.87% | 99.96% | 1.87% | 0.04 | | Diff | -8.47% | 0.40% | -8.47% | -0.15 | | low_vegetation | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 60.63% | 66.94% | 86.53% | 0.75 | | After | 55.99% | 58.07% | 94.01% | 0.72 | | Diff | -4.63% | -8.87% | 7.47% | -0.04 | | human_made_object | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 26.20% | 50.70% | 35.16% | 0.42 | | After | 0.00% | 0.00% | N/A | N/A | | Diff | -26.20% | -50.70% | N/A | N/A |
sheffield_park_small Overall accuracy from 91.66% to 91.37% | ground | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 91.12% | 92.07% | 98.89% | 0.95 | | After | 91.62% | 92.53% | 98.94% | 0.96 | | Diff | 0.50% | 0.46% | 0.05% | 0.00 | | building | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 81.49% | 83.50% | 97.12% | 0.90 | | After | 75.33% | 76.46% | 98.08% | 0.86 | | Diff | -6.16% | -7.05% | 0.95% | -0.04 | | low_vegetation | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 68.79% | 98.92% | 69.32% | 0.82 | | After | 80.03% | 94.96% | 83.58% | 0.89 | | Diff | 11.24% | -3.96% | 14.27% | 0.07 | | human_made_object | Accuracy | Recall | Precision | F1 | | -------------------- | -------- | -------- | --------- | -------- | | Before | 0.65% | 0.70% | 8.11% | 0.01 | | After | 12.76% | 86.17% | 13.03% | 0.23 | | Diff | 12.11% | 85.47% | 4.91% | 0.21 |

Average (2 datasets)

Overall accuracy from 91.96% to 77.55%

ground Accuracy Recall Precision F1
Before 91.79% 92.70% 98.94% 0.96
After 77.67% 78.32% 99.00% 0.87
Diff -14.11% -14.38% 0.06% -0.09
building Accuracy Recall Precision F1
Before 45.91% 91.53% 53.73% 0.54
After 38.60% 88.21% 49.97% 0.45
Diff -7.32% -3.32% -3.76% -0.09
low_vegetation Accuracy Recall Precision F1
Before 64.71% 82.93% 77.92% 0.78
After 68.01% 76.51% 88.79% 0.80
Diff 3.30% -6.42% 10.87% 0.02
human_made_object Accuracy Recall Precision F1
Before 13.43% 25.70% 21.64% 0.21
After 6.38% 43.08% 6.51% 0.11
Diff -7.04% 17.38% -15.12% -0.10