| """Read an extracted training sample. Needs h5py, numpy and Pillow only.""" |
| import argparse |
| import json |
| from pathlib import Path |
| import h5py |
| import numpy as np |
| from PIL import Image |
|
|
| parser=argparse.ArgumentParser() |
| parser.add_argument('sample_directory',type=Path) |
| parser.add_argument('--frame',type=int,default=0) |
| parser.add_argument('--height-index',type=int,default=0) |
| args=parser.parse_args();root=args.sample_directory |
| with h5py.File(root/'training.h5','r') as h: |
| rss_dbm=h['rf/rss_dbm'][args.frame,args.height_index] |
| valid=np.isfinite(rss_dbm) |
| timestamp=float(h['state/timestamps_s'][args.frame]) |
| height=float(h['state/height_levels_m'][args.height_index]) |
| sync=json.loads((root/'visual/synchronization.json').read_text()) |
| frame=sync['frames'][args.frame] |
| assert frame['timestamp_s']==timestamp |
| images={view:np.asarray(Image.open(root/'visual'/frame['cameras'][view]['rgb']).convert('RGB')) |
| for view in ['bs','topdown','elevated']} |
| print(json.dumps(dict(timestamp_s=timestamp,height_m=height,rss_shape=list(rss_dbm.shape), |
| unit='dBm',valid_fraction=float(valid.mean()), |
| rgb_shapes={view:list(img.shape) for view,img in images.items()}),ensure_ascii=False,indent=2)) |
|
|
