| """Validate the actual portable release with no processing-stack imports."""
|
| from pathlib import Path
|
| import sys,json,hashlib,math,subprocess,argparse,zipfile
|
| sys.dont_write_bytecode=True
|
| import numpy as np
|
| import polars as pl
|
| parser=argparse.ArgumentParser(description=__doc__);parser.add_argument("root",type=Path);parser.add_argument("--write-report",action="store_true")
|
| args=parser.parse_args();root=args.root.resolve();sys.path.insert(0,str(root))
|
| from observation_reader import ObservationDataset,camera_points
|
| from datasets import load_dataset,Video
|
| sample=ObservationDataset(root);report={"reader_environment":"isolated numpy/polars/opencv + Hugging Face Datasets", "episodes":[], "annotation_accuracy":"not_independently_measured", "robot_transfer":"not_evaluated"}
|
| assert len(sample.episodes)==3
|
| for key,e in sample.episodes.items():
|
| mapping=sample.table(key,"frame_mapping");display=mapping.filter(pl.col("rgb_decodable"))
|
| assert display.height==e["frame_count"]
|
| assert mapping["source_frame_index"].to_list()==list(range(mapping.height))
|
| with zipfile.ZipFile(root/e["sensors"]) as archive:
|
| names=set(archive.namelist())
|
| for source_idx in mapping["source_frame_index"]:
|
| assert all(f"{kind}/{source_idx:06d}.png" in names for kind in ("depth","confidence"))
|
| assert display["rgb_frame_index"].to_list()==list(range(e["frame_count"]))
|
| assert display["t_s"][0]==0
|
| indices=[0,e["frame_count"]//4,e["frame_count"]//2,3*e["frame_count"]//4,e["frame_count"]-1]
|
| tests=[]
|
| for idx in indices:
|
| frame=sample.frame(key,idx)
|
| assert frame["rgb"].shape==(e["height"],e["width"],3)
|
| assert frame["confidence"].shape==frame["depth_m"].shape
|
| assert np.isfinite(frame["K_rgb"]).all() and frame["K_rgb"][0,0]>0
|
| assert np.isfinite(frame["camera_pose_xyzw"]).all()
|
| assert abs(np.linalg.norm(frame["camera_pose_xyzw"][3:])-1)<1e-3
|
| assert (frame["depth_m"][~frame["depth_valid"]]==0).all()
|
| points=camera_points(frame); assert np.isfinite(points).all() and len(points)>0
|
| tests.append({"rgb_frame_index":idx,"source_frame_index":frame["source_frame_index"],"valid_depth_points":len(points)})
|
| hand=sample.table(key,"hand_pose")
|
| assert all(hand[k].null_count()==hand.height for k in ("z","wx","wy","wz"))
|
| assert sample.annotations_near(key,"hand_pose",1e6).height==0
|
| report["episodes"].append({"capture_id":key,"displayed_rgb_frames":e["frame_count"],"source_sensor_frames":mapping.height,"decoded_samples":tests,"hand_rows":hand.height,"handedness_rows":hand.group_by("handedness").len().sort("handedness").to_dicts()})
|
|
|
| catalog=load_dataset("parquet",data_files={"sample":str(root/"data/episodes.parquet")})["sample"]
|
| assert isinstance(catalog.features["video"],Video)
|
| catalog=catalog.cast_column("video",Video(decode=False));assert len(catalog)==3
|
| for row in catalog:
|
| assert row["video"]["bytes"] and row["instruction"]
|
| report["huggingface_preview"]={"rows":3,"video_feature":True,"embedded_video_bytes":True}
|
| checked=0
|
| for line in (root/"checksums.sha256").read_text().splitlines():
|
| expected,rel=line.split(" ",1);actual=hashlib.file_digest((root/rel).open("rb"),"sha256").hexdigest()
|
| assert actual==expected,rel
|
| checked+=1
|
| report["verified_artifact_hashes"]=checked
|
| report["status"]="passed"
|
| if args.write_report:
|
| (root/"validation.json").write_text(json.dumps(report,indent=2),encoding="utf-8")
|
| print(json.dumps(report,indent=2))
|
|
|
