| """EarthView (Satellogic 1 m, CC BY 4.0): build real same-place / different-date RGB pairs as 'no change' negatives. |
| |
| Downloads N shards, keeps locations with >= 2 revisits, stores up to 3 revisit pairs per location as uint8 |
| (K, 2, H, W, 3) at native 1 m. The trainer crops ~115 px and resizes to 192 (~0.6 m/px). |
| usage: python ev_prep.py N_SHARDS |
| """ |
| import os |
| import sys |
| from concurrent.futures import ThreadPoolExecutor |
|
|
| import numpy as np |
| import pyarrow.parquet as pq |
| from huggingface_hub import hf_hub_download |
|
|
| N = int(sys.argv[1]) if len(sys.argv) > 1 else 20 |
| OUT = "/workspace/prep/ev_imgs.npy" |
|
|
|
|
| def to_rgb8(a): |
| """(bands,H,W) raw -> (H,W,3) uint8 with a per-image percentile stretch (display-like).""" |
| a = np.asarray(a, np.float32)[:3].transpose(1, 2, 0) |
| lo, hi = np.percentile(a, 1), np.percentile(a, 99) |
| return np.clip((a - lo) * 255.0 / max(hi - lo, 1e-6), 0, 255).astype(np.uint8) |
|
|
|
|
| def shard(i): |
| p = hf_hub_download("satellogic/EarthView", f"satellogic/train-{i:05d}-of-07863.parquet", repo_type="dataset", |
| local_dir="/workspace/data/ev") |
| rng = np.random.default_rng(i) |
| out = [] |
| col = pq.read_table(p, columns=["rgb"])["rgb"].combine_chunks() |
| nrev = np.diff(col.offsets.to_numpy()) |
| vals = col.flatten().flatten().flatten().flatten().to_numpy(zero_copy_only=False).astype(np.uint8) |
| per = vals.size // max(nrev.sum(), 1) |
| side = int(round((per / 3) ** 0.5)) |
| imgs = vals.reshape(-1, 3, side, side) |
| start = np.concatenate([[0], np.cumsum(nrev)]) |
| pairs = [] |
| for j in range(len(nrev)): |
| x = to_rgb8(imgs[start[j]]) |
| if (x.max(2) > 0).mean() > 0.95 and x.std() > 8: |
| out.append(x) |
| R = nrev[j] |
| for _ in range(min(2, R - 1)): |
| a, b = rng.choice(R, 2, replace=False) |
| u, v = to_rgb8(imgs[start[j] + a]), to_rgb8(imgs[start[j] + b]) |
| if (u.max(2) > 0).mean() > 0.95 and (v.max(2) > 0).mean() > 0.95: |
| pairs.append(np.stack([u, v])) |
| print("shard", i, "rows", len(nrev), "multi-revisit", int((nrev >= 2).sum()), "max rev", int(nrev.max()), flush=True) |
| os.remove(p) |
| print("shard", i, len(out), flush=True) |
| return out, pairs |
|
|
|
|
| if __name__ == "__main__": |
| idx = np.random.default_rng(1).choice(7863, N, replace=False) |
| with ThreadPoolExecutor(8) as ex: |
| res = list(ex.map(shard, idx)) |
| singles = np.stack([x for o, _ in res for x in o if x.shape == (384, 384, 3)]) |
| pairs = [p for _, ps in res for p in ps if p.shape == (2, 384, 384, 3)] |
| np.save("/workspace/prep/ev_imgs.npy", singles) |
| np.save("/workspace/prep/ev_pairs.npy", np.stack(pairs) if pairs else np.zeros((0, 2, 384, 384, 3), np.uint8)) |
| print("EV_DONE singles", singles.shape, "pairs", len(pairs), flush=True) |
|
|
