Download read_example.py from KAS2003/xfield-sim-core: direct link, hf CLI and curl.
- Browser
- Download file 2.19 kB
-
https://huggingface.co/datasets/KAS2003/xfield-sim-core/resolve/main/read_example.py
- Command line
-
hf download hf://datasets/KAS2003/xfield-sim-core/read_example.py
-
curl -L -o read_example.py https://huggingface.co/datasets/KAS2003/xfield-sim-core/resolve/main/read_example.py
2.19 kB
| """Download one aligned training example and save a preview from its arrays.""" | |
| import argparse | |
| from pathlib import Path | |
| import numpy as np | |
| from PIL import Image, ImageDraw | |
| from huggingface_hub import HfApi, hf_hub_download | |
| REPO = 'KAS2003/xfield-sim-core' | |
| SAMPLE = 'demo/town12_fit.npz' | |
| def main(): | |
| parser = argparse.ArgumentParser() | |
| parser.add_argument('--revision', help='Dataset revision, defaults to the current release') | |
| parser.add_argument('--output', type=Path, default=Path('xfield_example.png')) | |
| args = parser.parse_args() | |
| rev = args.revision or HfApi(token=False).dataset_info(REPO, revision='refs/heads/main', expand=['sha']).sha | |
| file = hf_hub_download(REPO, SAMPLE, repo_type='dataset', revision=rev, token=False) | |
| with np.load(file, allow_pickle=False) as z: | |
| cube, rgb, K, T = (z[n] for n in ('cube','image_rgb','K','T')) | |
| assert cube.dtype == np.complex64 and cube.shape == (512,25,64) and np.isfinite(cube).all() and np.any(cube) | |
| assert rgb.shape == (3,360,640) and rgb.dtype == np.uint8 | |
| assert K.shape == (3,3) and T.shape == (4,4) and np.isfinite(K).all() and np.isfinite(T).all() | |
| assert bool(z['dense_present']) and all(n in z.files for n in ('role','xr_gt','static_y','dense_sem','dense_height','dense_lane')) | |
| print('key', str(z['key'])) | |
| for n in ('cube','image_rgb','K','T','role','xr_gt','static_y','dense_sem','dense_height','dense_lane'): | |
| print(n, z[n].shape, z[n].dtype) | |
| power = 10*np.log10(np.maximum((np.abs(cube)**2).sum(axis=2),1e-20)) | |
| lo, hi = np.percentile(power,[5,99]) | |
| ra = (255*np.clip((power-lo)/max(hi-lo,1e-6),0,1)).astype('uint8') | |
| camera = Image.fromarray(rgb.transpose(1,2,0)) | |
| radar = Image.fromarray(ra[::-1]).resize((192,360)).convert('RGB') | |
| out = Image.new('RGB',(832,388),'white') | |
| out.paste(camera,(0,28)); out.paste(radar,(640,28)) | |
| pen=ImageDraw.Draw(out);pen.text((7,7),'Ego RGB',fill='black');pen.text((648,7),'Radar power (range / angle)',fill='black') | |
| out.save(args.output) | |
| print('revision',rev) | |
| print('preview',args.output) | |
| if __name__=='__main__': | |
| main() | |