Repository navigation
Expand file tree
/
Copy pathstats.py
More file actions
46 lines (36 loc) · 1.53 KB
/
Copy pathstats.py
File metadata and controls
46 lines (36 loc) · 1.53 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
"""
Summarize per-sequence eval results from evaluate.py.
"""
import argparse
import json
import statistics as s
def hr(width=110):
print('─' * width)
def summarize(path, table=False):
results = json.load(open(path))
keys = ['ATE', 'ATE_H', 'ATE_V', 'RP_RMSE', 'drift_rate']
print(f"\n{path} ({len(results)} sequences)")
hr()
print(f"{'metric':>10s} {'mean':>8s} {'median':>8s} {'max':>8s} worst")
hr()
for k in keys:
vals = [r[k] for r in results]
worst = max(results, key=lambda r: r[k])
print(f"{k:>10s} {s.mean(vals):8.3f} {s.median(vals):8.3f} {max(vals):8.3f} {worst['name'][:50]}")
hr()
print(f"ATE > 10 m: {sum(1 for r in results if r['ATE'] > 10):3d} / {len(results)}")
print(f"drift_rate > 5%: {sum(1 for r in results if r['drift_rate'] > 5):3d} / {len(results)}")
if table:
print()
hr(95)
print(f"{'seq':50s} {'ATE':>8s} {'ATE_H':>8s} {'ATE_V':>8s} {'RTE':>8s} {'drift%':>8s}")
hr(95)
for r in sorted(results, key=lambda r: r['ATE'], reverse=True):
print(f"{r['name'][:50]:50s} {r['ATE']:8.3f} {r['ATE_H']:8.3f} {r['ATE_V']:8.3f} {r['RP_RMSE']:8.3f} {r['drift_rate']:8.2f}")
hr(95)
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--json', type=str, default='result/loss_result/result.json')
parser.add_argument('--table', action='store_true', help='print per-sequence table')
args = parser.parse_args()
summarize(args.json, table=args.table)