def main(args):
config = load_config(args.config)
check_channels(config)
check_classes(config)
assert torch.cuda.is_available(
), "No GPU support found. Check CUDA and NVidia Driver install."
assert torch.distributed.is_nccl_available(
), "No NCCL support found. Check your PyTorch install."
world_size = torch.cuda.device_count()
args.bs = args.bs if args.bs is not None else math.floor(os.cpu_count() /
world_size)
args.workers = args.workers if args.workers is not None else args.bs
palette, transparency = make_palette(
[classe["color"] for classe in config["classes"]])
args.cover = [
tile for tile in tiles_from_csv(os.path.expanduser(args.cover))
] if args.cover else None
args.out = os.path.expanduser(args.out)
log = Logs(os.path.join(args.out, "log"))
chkpt = torch.load(args.checkpoint, map_location=torch.device("cpu"))
chkpt["loader"] = "SemSeg"
log.log("neo predict on {} GPUs, with {} workers/GPU and {} tiles/batch".
format(world_size, args.workers, args.bs))
log.log("Model {} - UUID: {}".format(chkpt["nn"], chkpt["uuid"]))
log.log("---")
loader = load_module("neat_eo.loaders.{}".format(chkpt["loader"].lower()))
lock_file = os.path.abspath(os.path.join(args.out, str(uuid.uuid1())))
dataset = getattr(loader, chkpt["loader"])(
config,
chkpt["shape_in"][1:3],
args.dataset,
args.cover,
mode="predict",
metatiles=args.metatiles,
keep_borders=args.keep_borders,
)
mp.spawn(gpu_worker,
nprocs=world_size,
args=(world_size, lock_file, args, config, dataset, palette,
transparency))
if os.path.exists(lock_file):
os.remove(lock_file)
if not args.no_web_ui and dataset.cover:
template = "leaflet.html" if not args.web_ui_template else args.web_ui_template
base_url = args.web_ui_base_url if args.web_ui_base_url else "."
web_ui(args.out, base_url, dataset.cover, dataset.cover, "png",
template)
def main(args):
config = load_config(args.config)
args.out = os.path.expanduser(args.out)
args.cover = [
tile for tile in tiles_from_csv(os.path.expanduser(args.cover))
] if args.cover else None
if args.classes_weights:
try:
args.classes_weights = list(
map(float, args.classes_weights.split(",")))
except:
assert args.classes_weights == "auto", "invalid --classes_weights value"
else:
args.classes_weights = [
classe["weight"] for classe in config["classes"]
]
args.tiles_weights = ([(tile, weight) for tile, weight in tiles_from_csv(
os.path.expanduser(args.tiles_weights), extra_columns=True)]
if args.tiles_weights else None)
config["model"][
"loader"] = args.loader if args.loader else config["model"]["loader"]
config["model"]["ts"] = tuple(map(
int, args.ts.split(","))) if args.ts else config["model"]["ts"]
config["model"]["nn"] = args.nn if args.nn else config["model"]["nn"]
config["model"]["encoder"] = args.encoder if args.encoder else config[
"model"]["encoder"]
config["train"]["bs"] = args.bs if args.bs else config["train"]["bs"]
config["train"][
"loss"] = args.loss if args.loss else config["train"]["loss"]
config["train"]["optimizer"][
"name"] = args.optimizer if args.optimizer else config["train"][
"optimizer"]["name"]
config["train"]["optimizer"][
"lr"] = args.lr if args.lr else config["train"]["optimizer"]["lr"]
check_classes(config)
check_channels(config)
check_model(config)
log = Logs(os.path.join(args.out, "log"))
assert torch.cuda.is_available(
), "No GPU support found. Check CUDA and NVidia Driver install."
assert torch.distributed.is_nccl_available(
), "No NCCL support found. Check your PyTorch install."
world_size = torch.cuda.device_count() if args.train_dataset else 1
args.workers = min(
config["train"]["bs"] if not args.workers else args.workers,
math.floor(os.cpu_count() / world_size))
assert args.eval_dataset or args.train_dataset, "Provide at least one dataset"
if args.eval_dataset and not args.train_dataset and not args.checkpoint:
log.log(
"\n\nNOTICE: No Checkpoint provided for eval only. Seems peculiar.\n\n"
)
log.log("neo train/eval on {} GPUs, with {} workers/GPU".format(
world_size, args.workers))
log.log("---")
loader = load_module("neat_eo.loaders.{}".format(
config["model"]["loader"].lower()))
train_dataset = None
if args.train_dataset:
assert os.path.isdir(os.path.expanduser(
args.train_dataset)), "--train_dataset path is not a directory"
train_dataset = getattr(loader, config["model"]["loader"])(
config, config["model"]["ts"], args.train_dataset, args.cover,
args.tiles_weights, "train")
assert len(train_dataset), "Empty or Invalid --train_dataset content"
shape_in = train_dataset.shape_in
shape_out = train_dataset.shape_out
log.log("\nDataSet Training: {}".format(args.train_dataset))
if args.classes_weights == "auto":
args.classes_weights = compute_classes_weights(
args.train_dataset, config["classes"], args.cover,
os.cpu_count())
eval_dataset = None
if args.eval_dataset:
assert os.path.isdir(os.path.expanduser(
args.eval_dataset)), "--eval_dataset path is not a directory"
eval_dataset = getattr(loader, config["model"]["loader"])(
config, config["model"]["ts"], args.eval_dataset, args.cover,
args.tiles_weights, "eval")
assert len(eval_dataset), "Empty or Invalid --eval_dataset content"
shape_in = eval_dataset.shape_in
shape_out = eval_dataset.shape_out
log.log("DataSet Eval: {}".format(args.eval_dataset))
if not args.train_dataset and args.classes_weights == "auto":
args.classes_weights = compute_classes_weights(
args.eval_dataset, config["classes"], args.cover,
os.cpu_count())
log.log("\n--- Input tensor")
num_channel = 1 # 1-based numerotation
for channel in config["channels"]:
for band in channel["bands"]:
log.log("Channel {}:\t\t {} - (band:{})".format(
num_channel, channel["name"], band))
num_channel += 1
log.log("\n--- Output Classes ---")
for c, classe in enumerate(config["classes"]):
log.log("Class {}:\t\t {} ({:.2f})".format(c, classe["title"],
args.classes_weights[c]))
log.log("\n--- Model ---")
for hp in config["model"]:
log.log("{}{}".format(hp.ljust(25, " "), config["model"][hp]))
lock_file = os.path.abspath(os.path.join(args.out, str(uuid.uuid1())))
mp.spawn(
gpu_worker,
nprocs=world_size,
args=(world_size, lock_file, train_dataset, eval_dataset, shape_in,
shape_out, args, config),
)
if os.path.exists(lock_file):
os.remove(lock_file)
def gpu_worker(rank, world_size, lock_file, train_dataset, eval_dataset,
shape_in, shape_out, args, config):
log = Logs(os.path.join(args.out, "log")) if rank == 0 else None
csv_train = open(os.path.join(args.out, "train.csv"),
mode="a") if train_dataset and rank == 0 else None
csv_eval = open(os.path.join(args.out, "eval.csv"),
mode="a") if eval_dataset and rank == 0 else None
dist.init_process_group(backend="nccl",
init_method="file://" + lock_file,
world_size=world_size,
rank=rank)
torch.cuda.set_device(rank)
torch.manual_seed(0)
bs = config["train"]["bs"]
if train_dataset:
sampler = torch.utils.data.distributed.DistributedSampler(
train_dataset, num_replicas=world_size, rank=rank)
train_loader = DataLoader(train_dataset,
batch_size=bs,
shuffle=False,
drop_last=True,
num_workers=args.workers,
sampler=sampler)
else:
train_loader = None
if eval_dataset:
eval_loader = DataLoader(eval_dataset,
batch_size=bs,
shuffle=False,
drop_last=True,
num_workers=args.workers)
else:
eval_loader = None
nn_module = load_module("neat_eo.nn.{}".format(
config["model"]["nn"].lower()))
nn = getattr(nn_module,
config["model"]["nn"])(shape_in, shape_out,
config["model"]["encoder"].lower(),
config["train"]).cuda(rank)
nn = DistributedDataParallel(nn,
device_ids=[rank],
find_unused_parameters=True)
if train_dataset:
optimizer_params = {
key: value
for key, value in config["train"]["optimizer"].items()
if key != "name"
}
optimizer = getattr(torch.optim, config["train"]["optimizer"]["name"])(
nn.parameters(), **optimizer_params)
if rank == 0:
log.log("\n--- Train ---")
for hp in config["train"]:
if hp == "da":
da = config["train"]["da"]["name"]
dap = config["train"]["da"]["p"]
log.log("{}{} ({:.2f})".format("da".ljust(25, " "), da,
dap))
elif hp == "metrics":
log.log("{}{}".format(hp.ljust(
25, " "), set(config["train"][hp]))) # aesthetic
elif hp != "optimizer":
log.log("{}{}".format(hp.ljust(25, " "),
config["train"][hp]))
log.log("{}{}".format("optimizer".ljust(25, " "),
config["train"]["optimizer"]["name"]))
for k, v in optimizer.state_dict()["param_groups"][0].items():
if k != "params":
log.log(" - {}{}".format(k.ljust(25 - 3, " "), v))
resume = 0
if args.checkpoint:
chkpt = torch.load(os.path.expanduser(args.checkpoint),
map_location="cuda:{}".format(rank))
assert nn.module.version == chkpt[
"model_version"], "Model Version mismatch"
nn.load_state_dict(chkpt["state_dict"])
if rank == 0:
log.log("\n--- Using Checkpoint ---")
log.log("Path:\t\t {}".format(args.checkpoint))
log.log("UUID:\t\t {}".format(chkpt["uuid"]))
if args.resume:
optimizer.load_state_dict(chkpt["optimizer"])
resume = chkpt["epoch"]
assert resume < args.epochs, "Epoch asked, already reached by the given checkpoint"
loss_module = load_module("neat_eo.losses.{}".format(
config["train"]["loss"].lower()))
criterion = getattr(loss_module, config["train"]["loss"])().cuda(rank)
if eval_dataset and not train_dataset:
do_epoch(rank, eval_loader, config, args.classes_weights, log,
csv_eval, nn, criterion, "eval", 1)
dist.destroy_process_group()
return
for epoch in range(resume + 1, args.epochs + 1): # 1-N based
if train_dataset:
if rank == 0:
log.log("\n---\nEpoch: {}/{}\n".format(epoch, args.epochs))
sampler.set_epoch(
epoch) # https://github.com/pytorch/pytorch/issues/31232
do_epoch(rank, train_loader, config, args.classes_weights, log,
csv_train, nn, criterion, "train", epoch, optimizer)
if rank == 0:
UUID = uuid.uuid1()
states = {
"uuid": UUID,
"model_version": nn.module.version,
"producer_name": "Neat-EO.pink",
"producer_version": neo.__version__,
"model_licence": "MIT",
"domain": "pink.Neat-EO", # reverse-DNS
"doc_string": nn.module.doc_string,
"shape_in": shape_in,
"shape_out": shape_out,
"state_dict": nn.state_dict(),
"epoch": epoch,
"nn": config["model"]["nn"],
"encoder": config["model"]["encoder"],
"optimizer": optimizer.state_dict(),
"loader": config["model"]["loader"],
}
checkpoint_path = os.path.join(
args.out, "checkpoint-{:05d}.pth".format(epoch))
if epoch == args.epochs or not (epoch % args.saving):
log.log("\n--- Saving Checkpoint ---")
log.log("Path:\t\t {}".format(checkpoint_path))
log.log("UUID:\t\t {}\n".format(UUID))
torch.save(states, checkpoint_path)
dist.barrier()
if eval_dataset:
do_epoch(rank, eval_loader, config, args.classes_weights, log,
csv_eval, nn, criterion, "eval", epoch)
dist.destroy_process_group()
def main(args):
assert not (args.label and args.format), "Format option not supported for label, output must be kept as png"
try:
args.bands = list(map(int, args.bands.split(","))) if args.bands else None
except:
raise ValueError("invalid --args.bands value")
if not args.workers:
args.workers = min(os.cpu_count(), len(args.rasters))
if args.label:
config = load_config(args.config)
check_classes(config)
colors = [classe["color"] for classe in config["classes"]]
palette = make_palette(colors)
assert len(args.ts.split(",")) == 2, "--ts expect width,height value (e.g 512,512)"
width, height = list(map(int, args.ts.split(",")))
cover = [tile for tile in tiles_from_csv(os.path.expanduser(args.cover))] if args.cover else None
splits_path = os.path.join(os.path.expanduser(args.out), ".splits")
args.out = os.path.expanduser(args.out)
if os.path.dirname(os.path.expanduser(args.out)):
os.makedirs(args.out, exist_ok=True)
log = Logs(os.path.join(args.out, "log"), out=sys.stderr)
raster = rasterio_open(os.path.expanduser(args.rasters[0]))
args.bands = args.bands if args.bands else raster.indexes
raster.close()
print(
"neo tile {} rasters on bands {}, on CPU with {} workers".format(len(args.rasters), args.bands, args.workers),
file=sys.stderr,
flush=True,
)
skip = []
tiles_map = {}
total = 0
for path in args.rasters:
raster = rasterio_open(os.path.expanduser(path))
assert set(args.bands).issubset(set(raster.indexes)), "Missing bands in raster {}".format(path)
try:
w, s, e, n = transform_bounds(raster.crs, "EPSG:4326", *raster.bounds)
except:
log.log("WARNING: missing or invalid raster projection, SKIPPING: {}".format(path))
skip.append(path)
continue
tiles = [mercantile.Tile(x=x, y=y, z=z) for x, y, z in mercantile.tiles(w, s, e, n, args.zoom)]
tiles = list(set(tiles) & set(cover)) if cover else tiles
total += len(tiles)
for tile in tiles:
tile_key = (str(tile.x), str(tile.y), str(tile.z))
if tile_key not in tiles_map.keys():
tiles_map[tile_key] = []
tiles_map[tile_key].append(path)
raster.close()
assert total, "Nothing left to tile"
if len(args.bands) == 1 or args.label:
ext = "png" if args.format is None else args.format
if len(args.bands) == 3:
ext = "webp" if args.format is None else args.format
if len(args.bands) > 3:
ext = "tiff" if args.format is None else args.format
tiles = []
progress = tqdm(desc="Coverage tiling", total=total, ascii=True, unit="tile")
with futures.ThreadPoolExecutor(args.workers) as executor:
def worker(path):
if path in skip:
return None
raster = rasterio_open(path)
w, s, e, n = transform_bounds(raster.crs, "EPSG:4326", *raster.bounds)
tiles = [mercantile.Tile(x=x, y=y, z=z) for x, y, z in mercantile.tiles(w, s, e, n, args.zoom)]
tiled = []
for tile in tiles:
if cover and tile not in cover:
continue
w, s, e, n = mercantile.xy_bounds(tile)
warp_vrt = WarpedVRT(
raster,
crs="epsg:3857",
resampling=Resampling.bilinear,
add_alpha=False,
transform=from_bounds(w, s, e, n, width, height),
width=width,
height=height,
)
data = warp_vrt.read(
out_shape=(len(args.bands), width, height), indexes=args.bands, window=warp_vrt.window(w, s, e, n)
)
if data.dtype == "uint16": # GeoTiff could be 16 bits
data = np.uint8(data / 256)
elif data.dtype == "uint32": # or 32 bits
data = np.uint8(data / (256 * 256))
image = np.moveaxis(data, 0, 2) # C,H,W -> H,W,C
tile_key = (str(tile.x), str(tile.y), str(tile.z))
if (
not args.label
and len(tiles_map[tile_key]) == 1
and is_nodata(image, args.nodata, args.nodata_threshold, args.keep_borders)
):
progress.update()
continue
if len(tiles_map[tile_key]) > 1:
out = os.path.join(splits_path, str(tiles_map[tile_key].index(path)))
else:
out = args.out
x, y, z = map(int, tile)
if not args.label:
tile_image_to_file(out, mercantile.Tile(x=x, y=y, z=z), image, ext=ext)
if args.label:
tile_label_to_file(out, mercantile.Tile(x=x, y=y, z=z), palette, args.nodata, image)
if len(tiles_map[tile_key]) == 1:
tiled.append(mercantile.Tile(x=x, y=y, z=z))
progress.update()
raster.close()
return tiled
for tiled in executor.map(worker, args.rasters):
if tiled is not None:
tiles.extend(tiled)
total = sum([1 for tile_key in tiles_map.keys() if len(tiles_map[tile_key]) > 1])
progress = tqdm(desc="Aggregate splits", total=total, ascii=True, unit="tile")
with futures.ThreadPoolExecutor(args.workers) as executor:
def worker(tile_key):
if len(tiles_map[tile_key]) == 1:
return
image = np.zeros((width, height, len(args.bands)), np.uint8)
x, y, z = map(int, tile_key)
for i in range(len(tiles_map[tile_key])):
root = os.path.join(splits_path, str(i))
_, path = tile_from_xyz(root, x, y, z)
if not args.label:
split = tile_image_from_file(path)
if args.label:
split = tile_label_from_file(path)
if len(split.shape) == 2:
split = split.reshape((width, height, 1)) # H,W -> H,W,C
assert image.shape == split.shape, "{}, {}".format(image.shape, split.shape)
image[np.where(image == 0)] += split[np.where(image == 0)]
if not args.label and is_nodata(image, args.nodata, args.nodata_threshold, args.keep_borders):
progress.update()
return
tile = mercantile.Tile(x=x, y=y, z=z)
if not args.label:
tile_image_to_file(args.out, tile, image)
if args.label:
tile_label_to_file(args.out, tile, palette, image)
progress.update()
return tile
for tiled in executor.map(worker, tiles_map.keys()):
if tiled is not None:
tiles.append(tiled)
if splits_path and os.path.isdir(splits_path):
shutil.rmtree(splits_path) # Delete suffixes dir if any
if tiles and not args.no_web_ui:
template = "leaflet.html" if not args.web_ui_template else args.web_ui_template
base_url = args.web_ui_base_url if args.web_ui_base_url else "."
web_ui(args.out, base_url, tiles, tiles, ext, template)
def main(args):
assert not (args.geojson is not None and args.pg is not None), "You have to choose between --pg or --geojson"
assert len(args.ts.split(",")) == 2, "--ts expect width,height value (e.g 512,512)"
config = load_config(args.config)
check_classes(config)
args.pg = config["auth"]["pg"] if not args.pg and "pg" in config["auth"].keys() else args.pg
assert not (args.sql and not args.pg), "With --sql option, --pg dsn setting must also be provided"
palette, transparency = make_palette([classe["color"] for classe in config["classes"]], complementary=True)
index = [config["classes"].index(classe) for classe in config["classes"] if classe["title"] == args.type]
assert index, "Requested type is not contains in your config file classes."
burn_value = index[0]
assert 0 < burn_value <= 255
if args.sql:
assert "limit" not in args.sql.lower(), "LIMIT is not supported"
assert "TILE_GEOM" in args.sql, "TILE_GEOM filter not found in your SQL"
sql = re.sub(r"ST_Intersects( )*\((.*)?TILE_GEOM(.*)?\)", "1=1", args.sql, re.I)
assert sql and sql != args.sql, "Incorrect TILE_GEOM filter in your SQL"
if os.path.dirname(os.path.expanduser(args.out)):
os.makedirs(os.path.expanduser(args.out), exist_ok=True)
args.out = os.path.expanduser(args.out)
log = Logs(os.path.join(args.out, "log"), out=sys.stderr)
tiles = [tile for tile in tiles_from_csv(os.path.expanduser(args.cover))]
assert len(tiles), "Empty Cover: {}".format(args.cover)
if args.geojson:
zoom = tiles[0].z
assert not [tile for tile in tiles if tile.z != zoom], "Unsupported zoom mixed cover. Use PostGIS instead"
feature_map = collections.defaultdict(list)
log.log("neo rasterize - Compute spatial index")
for geojson_file in args.geojson:
with open(os.path.expanduser(geojson_file)) as geojson:
feature_collection = json.load(geojson)
srid = geojson_srid(feature_collection)
for i, feature in enumerate(tqdm(feature_collection["features"], ascii=True, unit="feature")):
feature_map = geojson_parse_feature(zoom, srid, feature_map, feature, args.buffer)
features = args.geojson
if args.sql:
conn = psycopg2.connect(args.pg)
db = conn.cursor()
db.execute("""SELECT ST_Srid("1") AS srid FROM ({} LIMIT 1) AS t("1")""".format(sql))
srid = db.fetchone()[0]
assert srid and int(srid) > 0, "Unable to retrieve geometry SRID."
features = args.sql
if not len(feature_map):
log.log("-----------------------------------------------")
log.log("NOTICE: no feature to rasterize, seems peculiar")
log.log("-----------------------------------------------")
log.log("neo rasterize - rasterizing {} from {} on cover {}".format(args.type, features, args.cover))
with open(os.path.join(os.path.expanduser(args.out), args.type.lower() + "_cover.csv"), mode="w") as cover:
for tile in tqdm(tiles, ascii=True, unit="tile"):
geojson = None
if args.sql:
w, s, e, n = tile_bbox(tile)
tile_geom = "ST_Transform(ST_MakeEnvelope({},{},{},{}, 4326), {})".format(w, s, e, n, srid)
query = """
WITH
sql AS ({}),
geom AS (SELECT "1" AS geom FROM sql AS t("1")),
json AS (SELECT '{{"type": "Feature", "geometry": '
|| ST_AsGeoJSON((ST_Dump(ST_Transform(ST_Force2D(geom.geom), 4326))).geom, 6)
|| '}}' AS features
FROM geom)
SELECT '{{"type": "FeatureCollection", "features": [' || Array_To_String(array_agg(features), ',') || ']}}'
FROM json
""".format(
args.sql.replace("TILE_GEOM", tile_geom)
)
db.execute(query)
row = db.fetchone()
try:
geojson = json.loads(row[0])["features"] if row and row[0] else None
except Exception:
log.log("Warning: Invalid geometries, skipping {}".format(tile))
conn = psycopg2.connect(args.pg)
db = conn.cursor()
if args.geojson:
geojson = feature_map[tile] if tile in feature_map else None
if geojson:
num = len(geojson)
out = geojson_tile_burn(tile, geojson, 4326, list(map(int, args.ts.split(","))), burn_value)
if not geojson or out is None:
num = 0
out = np.zeros(shape=list(map(int, args.ts.split(","))), dtype=np.uint8)
tile_label_to_file(args.out, tile, palette, transparency, out, append=args.append)
cover.write("{},{},{} {}{}".format(tile.x, tile.y, tile.z, num, os.linesep))
if not args.no_web_ui:
template = "leaflet.html" if not args.web_ui_template else args.web_ui_template
base_url = args.web_ui_base_url if args.web_ui_base_url else "."
tiles = [tile for tile in tiles_from_csv(args.cover)]
web_ui(args.out, base_url, tiles, tiles, "png", template)
def main(args):
tiles = list(tiles_from_csv(args.cover))
assert len(tiles), "Empty cover: {}".format(args.cover)
args.workers = min(os.cpu_count(),
args.rate) if not args.workers else args.workers
if os.path.dirname(os.path.expanduser(args.out)):
os.makedirs(os.path.expanduser(args.out), exist_ok=True)
log = Logs(os.path.join(args.out, "log"), out=sys.stderr)
log.log("neo download with {} workers, at max {} req/s, from: {}".format(
args.workers, args.rate, args.url))
already_dl = 0
dl = 0
with requests.Session() as session:
progress = tqdm(total=len(tiles), ascii=True, unit="image")
with futures.ThreadPoolExecutor(args.workers) as executor:
def worker(tile):
tick = time.monotonic()
progress.update()
try:
x, y, z = map(str, [tile.x, tile.y, tile.z])
os.makedirs(os.path.join(args.out, z, x), exist_ok=True)
except:
return tile, None, False
path = os.path.join(args.out, z, x,
"{}.{}".format(y, args.format))
if os.path.isfile(path): # already downloaded
return tile, None, True
if args.type == "XYZ":
url = args.url.format(x=tile.x, y=tile.y, z=tile.z)
elif args.type == "WMS":
xmin, ymin, xmax, ymax = xy_bounds(tile)
url = args.url.format(xmin=xmin,
ymin=ymin,
xmax=xmax,
ymax=ymax)
res = tile_image_from_url(session, url, args.timeout)
if res is None: # let's retry once
res = tile_image_from_url(session, url, args.timeout)
if res is None:
return tile, url, False
try:
tile_image_to_file(args.out, tile, res)
except OSError:
return tile, url, False
tock = time.monotonic()
time_for_req = tock - tick
time_per_worker = args.workers / args.rate
if time_for_req < time_per_worker:
time.sleep(time_per_worker - time_for_req)
return tile, url, True
for tile, url, ok in executor.map(worker, tiles):
if url and ok:
dl += 1
elif not url and ok:
already_dl += 1
else:
log.log("Warning:\n {} failed, skipping.\n {}\n".format(
tile, url))
if already_dl:
log.log(
"Notice: {} tiles were already downloaded previously, and so skipped now."
.format(already_dl))
if already_dl + dl == len(tiles):
log.log("Notice: Coverage is fully downloaded.")
if not args.no_web_ui:
template = "leaflet.html" if not args.web_ui_template else args.web_ui_template
base_url = args.web_ui_base_url if args.web_ui_base_url else "."
web_ui(args.out, base_url, tiles, tiles, args.format, template)
def main(args):
assert args.cover or args.granules or args.scenes, "Either --cover OR --granules OR --scenes is mandatory"
assert not (args.download
and not args.out), "--download implies out parameter"
assert args.limit, "What about increasing --limit value ?"
config = load_config(args.config)
if args.cover:
args.pg = args.pg if args.pg else config["auth"]["pg"]
assert args.pg, "PostgreSQL connection settting is mandatory with --cover"
args.granules = tiles_to_granules(
tiles_from_csv(os.path.expanduser(args.cover)), args.pg)
if args.out:
args.out = os.path.expanduser(args.out)
os.makedirs(args.out, exist_ok=True)
log = Logs(os.path.join(args.out, "log"), out=sys.stderr)
else:
log = Logs(None, out=sys.stderr)
log.log("neo sat on granules: {}".format(" ".join(args.granules)))
scenes = search_scenes(args, log)
if args.download:
log.log("")
log.log(
"============================================================================="
)
log.log("Downloading selected scenes")
log.log(
"============================================================================="
)
report = []
login, password = dict([
auth.split("=") for auth in config["auth"]["theia"].split(" ")
]).values()
with futures.ThreadPoolExecutor(args.workers) as executor:
def worker(scene):
scene_dir = os.path.join(
args.out, scene["dir"]
[:42]) # 42 related to Theia MD issue, dirty workaround
if not os.path.isabs(scene_dir):
scene_dir = "./" + scene_dir
if glob.glob(scene_dir + "*"):
scene["dir"] = glob.glob(scene_dir + "*")[0]
return scene, None, True # Already Downloaded
token = get_token(login, password)
url = THEIA_URL + "/resto2/collections/SENTINEL2/{}/download/?issuerId=theia".format(
scene["uuid"])
resp = requests.get(
url,
headers={"Authorization": "Bearer {}".format(token)},
stream=True)
if resp is None:
return scene, None, False # Auth issue
zip_path = os.path.join(args.out, scene["uuid"] + ".zip")
with open(zip_path, "wb") as fp:
progress = tqdm(unit="B",
desc=scene["uuid"],
total=int(resp.headers["Content-Length"]))
for chunk in resp.iter_content(chunk_size=16384):
progress.update(16384)
fp.write(chunk)
return scene, zip_path, True
return scene, None, False # Write issue
for scene, zip_path, ok in executor.map(worker, scenes):
if zip_path and md5(zip_path) == scene["checksum"]:
scene["dir"] = os.path.dirname(
ZipFile(zip_path).namelist()[0])
ZipFile(zip_path).extractall(args.out)
os.remove(zip_path)
report.append("Scene {} available in {}".format(
scene["uuid"], scene["dir"]))
elif ok:
report.append(
"SKIPPING downloading {}, as already in {}".format(
scene["uuid"], scene["dir"]))
else:
report.append("ERROR: Unable to retrieve Scene {}".format(
scene["uuid"]))
log.log("")
log.log(
"============================================================================="
)
for line in report:
log.log(line)
log.log(
"============================================================================="
)