def map(
imbin_file,
som_file,
map_file,
trans_file,
som_width,
som_height,
cpu_only=False,
nrot=360,
numthreads=10,
overwrite=False,
log=True,
):
# /// Map through SOM \\\
imgs = pu.ImageReader(imbin_file)
if not os.path.exists(map_file) or overwrite:
map_imbin(
imbin_file,
som_file,
map_file,
trans_file,
som_width,
som_height,
numthreads=numthreads,
cpu=cpu_only,
nrot=nrot,
log=log,
)
som = pu.SOM(som_file)
mapping = pu.Mapping(map_file)
transform = pu.Transform(trans_file)
somset = pu.SOMSet(som, mapping, transform)
return somset
def preprocess(
radio_cat,
imbin_file,
img_size=(2, 150, 150),
tile_cutout_path="",
remove_tile_cutouts=False,
overwrite=False,
):
# /// Preprocess Data \\\
if not os.path.exists(imbin_file) or overwrite:
preprocessing.main(radio_cat, imbin_file, img_size, tile_cutout_path)
if remove_tile_cutouts:
shutil.rmtree(tile_cutout_path)
imgs = pu.ImageReader(imbin_file)
return imgs
return gaussian_2d(X, Y, 1, img_x / 2, img_y / 2, sig_x, sig_y, 0)
def taper_img(img, **kwargs):
return img * kernel(img.shape, **kwargs)
def reweight(data, old_weights, new_weights):
data = np.array(data)
for chan in range(data.shape[1]):
data[:, chan] *= new_weights[chan] / old_weights[chan]
return data
imgs = pu.ImageReader(
"EMU_WISE_E95E05_Aegean_Components_Complex_EMUWISE_IslandNorm_Log_Reprojected.bin"
)
# data = imgs.reweight(old_weights=(0.95, 0.05), new_weights=(1, 1))
data = reweight(imgs.data, (0.95, 0.05), (1, 1))
data = taper_img(data, fwhm_frac=0.5)
i = np.random.randint(imgs.data.shape[0])
fig, axes = plt.subplots(2, 2, figsize=(8, 8))
axes[0, 0].imshow(imgs.data[i, 0])
axes[0, 1].imshow(imgs.data[i, 1])
axes[1, 0].imshow(data[i, 0])
axes[1, 1].imshow(data[i, 1])
rad_tot = data[:, 0].reshape(-1, 150 * 150).sum(axis=1)
ir_tot = data[:, 1].reshape(-1, 150 * 150).sum(axis=1)
som_width,
som_height,
numthreads=cpu_count(),
cpu=args.cpu,
nrot=360,
log=True,
)
else:
print(
f"Mapping binary {map_file} and Transform file {trans_file} already exist...skipping."
)
# CONTINUE HERE if Mapping was done on a different machine.
# Update the component catalogue with the sidelobe probability
imgs = pu.ImageReader(imbin_file)
# Output a table of components that failed to preprocess
# This is usually due to missing files
failed = sample.iloc[list(set(sample.index).difference(
imgs.records))].reset_index(drop=True)
Table.from_pandas(failed).write(f"{cat_name}_failed.fits", overwrite=True)
# Output a table of all preprocessed components.
# This matches the length of the IMG and MAP binaries, which makes
# it useful for inspecting the results.
sample = sample.iloc[imgs.records].reset_index(drop=True)
Table.from_pandas(sample).write(f"{cat_name}_preprocessed.fits",
overwrite=True)
del imgs
print(f" \centering", file=outfile)
print(f" \includegraphics[width=0.75\\textwidth]{{{base}.png}}",
file=outfile)
print(f" \includegraphics[width=0.75\\textwidth]{{{base}_img0.png}}",
file=outfile)
print(f" \includegraphics[width=0.75\\textwidth]{{{base}_img1.png}}",
file=outfile)
print(f" \caption{{{caption}}}", file=outfile)
print(f" \label{{fig:{label}}}", file=outfile)
print(f"\end{{figure*}}", file=outfile)
if len(sys.argv) < 5:
print(f"USAGE: {sys.argv[0]} IMG SOM MAP TRANSFORM")
imgs = pu.ImageReader(sys.argv[1])
som = pu.SOM(sys.argv[2])
mapping = pu.Mapping(sys.argv[3])
trans = pu.Transform(sys.argv[4])
somset = pu.SOMSet(som, mapping, trans)
path = pu.PathHelper("Inspection")
outfile = open("latex_imgs.tex", "w")
### SOM 2D histogram ###
som_counts(somset, True)
plt.savefig(f"som_stats.png")
plt.clf()
caption = f"The total number of images within the {imgs.data.shape[0]} input images mapped to each neuron in the SOM."
latex_figure("som_stats.png", caption, "som_stats", outfile)
print("", file=outfile)
axes[2].imshow(somset.som[bmu_idx][0])
axes[2].set_title(f"Best-matching neuron: {bmu_idx}", fontsize=12)
# axes[2].imshow(somset.som[bmu_idx][1])
if df is not None:
fig.suptitle(df.loc[idx]["Component_name"], fontsize=16)
if len(sys.argv) != 3:
print("USAGE: {} sample_file som_file".format(sys.argv[0]))
sys.exit(-1)
sample_file = sys.argv[1]
df = pd.read_csv(sample_file)
imbin = pu.ImageReader(sample_file.replace(".csv", "_imgs.bin"))
som_file = sys.argv[2]
som = pu.SOM(som_file)
map_file = som_file.replace(".bin", "_Similarity.bin")
mapping = pu.Mapping(map_file)
trans_file = som_file.replace(".bin", "_Transform.bin")
transform = pu.Transform(trans_file)
somset = pu.SOMSet(som=som, mapping=mapping, transform=transform)
### Update the component catalogue ###
bmus = mapping.bmu()
def source_from_catalogue(
src_cat,
radio_cat,
imbin_path,
som=None,
show_nearby=False,
show_bmu=False,
idx=None,
):
if idx is None:
# idx = np.random.randint(len(src_cat))
idx = src_cat[src_cat.N_components > 1].sample(1).index[0]
src = src_cat.loc[idx]
comp_names = src.Component_names.split(";")
tile_id = radio_cat[radio_cat.Component_name == comp_names[0]].Tile.iloc[0]
radio_tile = radio_cat[radio_cat.Tile == tile_id].reset_index(drop=True)
comps = radio_tile[radio_tile["Component_name"].isin(comp_names)]
radio_ra = np.array(src.RA_components.split(";"), dtype=float)
radio_dec = np.array(src.DEC_components.split(";"), dtype=float)
radio_pos = SkyCoord(radio_ra, radio_dec, unit=u.deg)
if src.N_host_candidates > 0:
ir_ra = np.array(src.RA_host_candidates.split(";"), dtype=float)
ir_dec = np.array(src.DEC_host_candidates.split(";"), dtype=float)
ir_pos = SkyCoord(ir_ra, ir_dec, unit=u.deg)
img_file, map_file, trans_file = binary_names(tile_id, imbin_path)
imgs = pu.ImageReader(img_file)
mapping = pu.Mapping(map_file)
transform = pu.Transform(trans_file)
if som is not None:
somset = pu.SOMSet(som, mapping, transform)
img_idx = comps.index[0]
comp = comps.loc[img_idx]
npix = imgs.data.shape[-1]
wcs = create_wcs(comp.RA, comp.DEC, npix, 3 * u.arcmin / npix)
if show_bmu:
plot_image(
imgs,
img_idx,
somset=somset,
wcs=wcs,
transform_neuron=True,
show_bmu=show_bmu,
)
else:
plot_image(imgs, img_idx, wcs=wcs, show_bmu=False)
axes = plt.gcf().axes
if show_nearby:
posn = SkyCoord(comp.RA, comp.DEC, unit=u.deg)
coords = SkyCoord(radio_cat.RA, radio_cat.DEC, unit=u.deg)
nearby = coords[posn.separation(coords) < 1.5 * u.arcmin]
for ax in plt.gcf().axes:
ax.scatter(
nearby.RA, nearby.DEC, c="w", transform=ax.get_transform("world")
)
# for ax in plt.gcf().axes:
# ax.scatter(comps.RA, comps.DEC, c="r", transform=ax.get_transform("world"))
axes[0].scatter(
radio_pos.ra, radio_pos.dec, c="r", transform=axes[0].get_transform("world")
)
if src.N_host_candidates > 0:
axes[1].scatter(
ir_pos.ra, ir_pos.dec, c="w", transform=axes[1].get_transform("world")
)
plt.suptitle(f"Source ID: {idx}")
pandas=True)
# Sample of pairs
sample_path = "/home/adrian/CIRADA/Sample"
sample_file = "Pairs_uniformNNdist_5asecbins_minSN10_cleansep.csv"
radio_cat = vdl.load_vlass_catalogue(
os.path.join(sample_path, sample_file),
pandas=True,
complex=False,
)
ir_file = "Pairs_uniformNNdist_5asecbins_minSN10_cleansep_IR.fits"
ir_cat = load_unwise(ir_file, path=sample_path)
imbin_file = "IMG_uniformNNdist_5asecbins_minSN10_cleansep_compfilt_w7525.bin"
imgs = pu.ImageReader(os.path.join(sample_path, imbin_file))
# SOM Input Info
som_path = os.path.join("/home/adrian/CIRADA/SOM", "Pairs",
"ComponentFiltered", "w7525")
som_file = os.path.join(som_path, "SOM_B3_h10_w10_vlass.bin")
map_file = "MAP_uniformNNdist_5asecbins_minSN10_cleansep_compfilt_w7525_ComponentFiltered_10x10.bin"
trans_file = map_file.replace("MAP", "TRANSFORM")
somset = pu.SOMSet(
som_file,
os.path.join(sample_path, map_file),
os.path.join(sample_path, trans_file),
)
somset.som.bmu_mask = None
default="Plots",
type=str,
)
args = parser.parse_args()
return args
if __name__ == "__main__":
args = parse_args()
# Initialize the sample DataFrame, ImageReader, and SOM
# files = set_filenames(args.som_file, args.map_file, args.trans_file, args.outbase)
# df = pd.read_csv(args.sample_file)
imgs = pu.ImageReader(args.imbin_file)
somset = init_somset(args.som_file, args.map_file, args.trans_file)
path = pu.PathHelper(args.outpath)
# Only keep the records that preprocessed successfully
# df = df.loc[imgs.records].reset_index(drop=True)
# df = update_component_catalogue(df, somset)
som_counts(somset, show_counts=True)
plt.savefig(f"{path}/bmu_frequency.png")
# plot_image(imgs, df=df, somset=somset, show_bmu=True, apply_transform=True)
# dist_hist(df, labels=True)
# plt.savefig(f"{path}/euc_distance_hist.png")
# dist_hist_2d(df, somset, bins=100, loglog=False)