File size: 1,705 Bytes
aaa8065 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 | import sys,json
from pathlib import Path
sys.path.insert(0,str(Path(__file__).resolve().parents[1]))
import numpy as np
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
from dataset import get_meta,train_patch_ids
p=Path(__file__).parent;d=json.loads((p/'plateau_diagnostics.json').read_text());test=json.loads(Path('analysis_oct05/per_class_audit.json').read_text())
plt.rcParams.update({'font.size':10,'axes.spines.top':False,'axes.spines.right':False})
fig,axs=plt.subplots(1,2,figsize=(13,5),layout='constrained')
seq=sorted(set(tuple(get_meta()[pid]['dates']) for pid in train_patch_ids()))
for k,days in enumerate(seq):axs[0].plot(np.arange(len(days)),days,label=f'Calendar {k+1}')
axs[0].axvline(20,color='grey',linestyle=':',alpha=.7)
axs[0].set(xlabel='Acquisition index supplied to the model',ylabel='Actual day offset in metadata',title='The same index can mean different seasonal dates')
axs[0].legend(loc='lower right',fontsize=9)
classes=[17,10,18,13,12,14,9];pos=np.arange(len(classes))
for label,key,color,offset in [('Validation','sam_val_iou','#2577ad',-.18),('Test','test_iou','#d88b33',.18)]:
values=[100*(d['per_class_fixed_validation'][c-1][key] if label=='Validation' else test[c-1][key]) for c in classes]
axs[1].barh(pos+offset,values,height=.34,label=label,color=color)
labels=[d['per_class_fixed_validation'][c-1]['name']+f" ({d['per_class_fixed_validation'][c-1]['train_unique_patches']} train patches)" for c in classes]
axs[1].set_yticks(pos,labels);axs[1].invert_yaxis();axs[1].set(xlim=(0,100),xlabel='Class IoU (%)',title='Rare-class evidence is geographically sparse');axs[1].legend()
fig.savefig(p/'plateau_evidence.png',dpi=180);plt.close(fig)
|