dfichuk/tree-species-high-accuracy-psp

High-Accuracy LiDAR-backed PSP Dataset

High-accuracy PSP subset with LiDAR-backed patches.

Size

• —5,737 LiDAR-backed high-coordinate-accuracy PSP tree samples

Species counts

• —BA: 36
• —CW: 1,362
• —DR: 557
• —FD: 651
• —HW: 2,776
• —MB: 52
• —SS: 303

Coordinate reliability filtering

Coordinate tiers follow the source PSP coordinate provenance:

• —High: DGPS, PPP
• —Medium: RGPS, VGPS
• —Low: MAP, GIS, GE, PREVIOUS, INTENDED, UNKNOWN, missing, or unrecognized

This subset keeps only rows with high-coordinate-accuracy labels.

Label provenance and reliability

• —Species labels: professional PSP field labels
• —DBH: field measured
• —Height: mostly estimated from DBH equations unless height_source says Field measured
• —LiDAR patches: quality-filtered 5m-radius point clouds

Split counts

• —train: 3,938
• —validation: 900
• —test: 899

Patch format

Each patch is stored as a .npy array with per-point fields:

• —x
• —y
• —z
• —intensity
• —return_number
• —number_of_returns
• —classification

Loading example

import pandas as pd
import numpy as np
from pathlib import Path

root = Path("./huggingface_high_accuracy_lidar")
df = pd.read_csv(root / "high_accuracy_lidar_set.csv")
row = df.iloc[0]
patch = np.load(root / row.patch_file, allow_pickle=False)
print(row.species, patch.dtype)

Caveats

• —Plot coordinates still reflect PSP plot/tree matching limitations.
• —Only DGPS/PPP is included in this release.
• —Patches are 5m radius.
• —No interrupted download/extraction continuation is included.