#!/usr/bin/env python3 """Everything a sitting photographed, as one folder a splat can be trained from. python scan_view_folder.py [--out ] Print What I See keeps two kinds of pictures: the scan's (scan/k0000.jpg ... with depth and a row in scan/frames.csv) and one per trigger pull (capture_.jpg with capture_.json: the camera pose at that moment). This writes /view/ with all of them under one frames.csv and the scan's camera.json, so quest_scan.py --scan <...>/view trains one splat of everything that was looked at. There is no scan.json in it: it is a view of the surroundings, not one object. """ import argparse import csv import json import os import shutil def main(): ap = argparse.ArgumentParser() ap.add_argument("folder") ap.add_argument("--out") a = ap.parse_args() src = os.path.normpath(a.folder) scan = os.path.join(src, "scan") out = os.path.normpath(a.out) if a.out else os.path.join(src, "view") shutil.rmtree(out, ignore_errors=True) os.makedirs(out) cam = os.path.join(scan, "camera.json") if not os.path.exists(cam): raise SystemExit(f"no scan/camera.json in {src}: this sitting has no scan pictures to take the camera from") shutil.copyfile(cam, os.path.join(out, "camera.json")) rows = [] for r in csv.DictReader(open(os.path.join(scan, "frames.csv"))): p = os.path.join(scan, r["file"]) if not os.path.exists(p): continue shutil.copyfile(p, os.path.join(out, r["file"])) d = os.path.splitext(p)[0] + ".depth" if os.path.exists(d): shutil.copyfile(d, os.path.join(out, os.path.basename(d))) rows.append({k: r[k] for k in ("file", "t_unix_ms", "t_app_s", "cam_px", "cam_py", "cam_pz", "cam_qx", "cam_qy", "cam_qz", "cam_qw")}) n_scan = len(rows) anchors = {} caps = sorted((f for f in os.listdir(src) if f.startswith("capture_") and f.endswith(".json")), key=lambda s: int(s[8:-5])) for c in caps: j = json.load(open(os.path.join(src, c))) jpg = os.path.join(src, c[:-5] + ".jpg") if not os.path.exists(jpg): continue name = "p%04d.jpg" % int(c[8:-5]) shutil.copyfile(jpg, os.path.join(out, name)) p, q = j["cam_pos"], j["cam_rot"] rows.append({"file": name, "t_unix_ms": j.get("t_unix_ms", 0), "t_app_s": 0, "cam_px": p[0], "cam_py": p[1], "cam_pz": p[2], "cam_qx": q[0], "cam_qy": q[1], "cam_qz": q[2], "cam_qw": q[3]}) if j.get("seed"): anchors[name] = j["seed"] # where the trigger pointed: the one spot of this picture with a measured depth json.dump({"coordinates": "Unity world, left-handed, y up, metres", "points": anchors}, open(os.path.join(out, "anchors.json"), "w"), indent=1) with open(os.path.join(out, "frames.csv"), "w", newline="") as fh: w = csv.DictWriter(fh, fieldnames=["frame", "file", "t_unix_ms", "t_app_s", "cam_px", "cam_py", "cam_pz", "cam_qx", "cam_qy", "cam_qz", "cam_qw"]) w.writeheader() for i, r in enumerate(rows): w.writerow(dict(r, frame=i)) print(f"{out}: {n_scan} scan pictures + {len(rows) - n_scan} trigger pictures") if __name__ == "__main__": main()