grid[13].set_xlabel('feat = 5K')
grid[14].set_xlabel('feat = 10K')
grid[15].set_xlabel('feat = 20K')
plt.savefig('/Users/venabled/doc/journal1/img/neogeo_grid_out.png', dpi=600)
# Get a truth finder
f5 = tb.open_file('/Users/venabled/data/uvan/fc2_f5.hdf', 'r')
imgs = f5.root.camera.image_raw.compressed.images
img_times = imgs = f5.root.camera.image_raw.compressed.metadata.cols.t_valid
pva = f5.root.nov_span.pva
srtm_path = '/Users/venabled/data/srtm/SRTM1/Region_01'
geoid_file = '/Users/venabled/data/geoid/egm96_15.tiff'
frame_yaml = '/Users/venabled/pysrc/pnpnav/data/fc2_pod_frames.yaml'
finder = pnputils.DEMTileFinder(srtm_path, geoid_file)
finder.load_cam_and_vehicle_frames(frame_yaml)
finder.load_camera_cal('/Users/venabled/pysrc/pnpnav/data/fc2_cam_model.yaml')
corners = np.zeros((img_times.shape[0], 4, 3))
for img_num in np.arange(img_times.shape[0]):
dt = pva.cols.t_valid - img_times[img_num]
ii = np.abs(dt).argmin()
lon_lat_h = np.array([pva[ii]['lon'], pva[ii]['lat'], pva[ii]['height']])
att = pva[ii]['c_nav_veh'].reshape(3, 3)
corners[img_num] = finder.get_corners(lon_lat_h, att)[1]
print('image: %d / %d' % (img_num, img_times.shape[0]))
centers = np.array([corner[:, :2].mean(0) for corner in corners])
bbox = neoextent.bbox_from_extent(extent)
c_wgs = np.hstack((np.array(bbox)[:, [1, 0]], np.zeros((4, 1))))
# obs_out[img_num] = np.copy(obs_likelihood)
if __name__ == '__main__':
f5 = tb.open_file('/media/sean/D2F2E7B2F2E798CD/Users/student/neo_data/fc2_f5.hdf', 'r')
f5featmeta = pd.read_hdf('/media/sean/D2F2E7B2F2E798CD/Users/student/AIEoutput2/feat/feat_meta.hdf')
img_times = f5.root.camera.image_raw.compressed.metadata.col('t_valid')
feat_path = f5featmeta.iloc[:, 3]
descripath = f5featmeta.iloc[:, 4]
# Get a truth finder
dted_path = '/media/sean/D2F2E7B2F2E798CD/Users/student/neo_data/srtm'
geoid_file = '/media/sean/D2F2E7B2F2E798CD/Users/student/neo_data/egm96-15.tif'
frame_yaml = '/home/sean/ImageAidedNav/pypnp/data/fc2_pod_frames.yaml'
finder = pnputils.DEMTileFinder(dted_path, geoid_file)
finder.load_cam_and_vehicle_frames(frame_yaml)
finder.load_camera_cal('/home/sean/ImageAidedNav/pypnp/data/fc2_cam_model.yaml')
# Get the Feature Database
dbf = tb.open_file('/media/sean/D2F2E7B2F2E798CD/Users/student/neo_data/pytables_db.hdf', 'r')
dbt = dbf.get_node('/sift_db/sift_features_sorted')
# Set up the output file
out_tb = tb.open_file('/media/sean/D2F2E7B2F2E798CD/Users/student/neo_data/obs_out_mat7.hdf', 'w')
# You need boundaries
txmin = dbt.cols.x[dbt.colindexes['x'][0]].astype(np.int64)
txmax = dbt.cols.x[dbt.colindexes['x'][-1]].astype(np.int64)
tymin = dbt.cols.y[dbt.colindexes['y'][0]].astype(np.int64)
tymax = dbt.cols.y[dbt.colindexes['y'][-1]].astype(np.int64)
def FMOTF(
f5,
f5featmeta,
dted_path,
geoid_file,
frame_yaml,
cam_yaml,
dbf,
out_path,
):
# f5 = tb.open_file('/media/sean/D2F2E7B2F2E798CD/Users/student/neo_data/fc2_f5.hdf', 'r')
# f5featmeta = pd.read_hdf('/media/sean/D2F2E7B2F2E798CD/Users/student/AIEoutput2/feat/feat_meta.hdf')
img_times = f5.root.camera.image_raw.compressed.metadata.col('t_valid')
feat_path = f5featmeta.iloc[:, 3]
descripath = f5featmeta.iloc[:, 4]
# Get a truth finder
# dted_path = '/media/sean/D2F2E7B2F2E798CD/Users/student/neo_data/srtm'
# geoid_file = '/media/sean/D2F2E7B2F2E798CD/Users/student/neo_data/egm96-15.tif'
# frame_yaml = '/home/sean/ImageAidedNav/pypnp/data/fc2_pod_frames.yaml'
# cam_yaml = '/home/sean/ImageAidedNav/pypnp/data/fc2_cam_model.yaml'
finder = pnputils.DEMTileFinder(dted_path, geoid_file)
finder.load_cam_and_vehicle_frames(frame_yaml)
finder.load_camera_cal(cam_yaml)
# Get the Feature Database
# dbf = tb.open_file('/media/sean/D2F2E7B2F2E798CD/Users/student/neo_data/pytables_db.hdf', 'r')
dbt = dbf.get_node('/sift_db/sift_features_sorted')
# Set up the output file
# out_path = '/media/sean/D2F2E7B2F2E798CD/Users/student/neo_data/obs_out_mat7.hdf'
out_tb = tb.open_file(out_path, 'w')
# You need boundaries
txmin = dbt.cols.x[dbt.colindexes['x'][0]].astype(np.int64)
txmax = dbt.cols.x[dbt.colindexes['x'][-1]].astype(np.int64)
tymin = dbt.cols.y[dbt.colindexes['y'][0]].astype(np.int64)
tymax = dbt.cols.y[dbt.colindexes['y'][-1]].astype(np.int64)
xbounds = (txmin, txmax)
ybounds = (tymin, tymax)
print(xbounds)
print(ybounds)
xb, yb = neoextent.pad_grid(xbounds, ybounds)
# TODO Redefine xb and yb to be the tiles that encompassed the area defined by the particles and/or confidence ellipse
f2l_mat = np.array([1E6])
n_img_feat = np.array([10E3])
# Create blessings.Terminal instance
leaf_values, test_d = build_progress(f2l_mat, n_img_feat, img_times)
tt = Terminal()
prog_tree = ProgressTree(term=tt)
prog_tree.make_room(test_d)
# Full aggregate tile id
tid, tidcount = np.unique(dbt.cols.pair_id, return_counts=True)
extent = neoextent.SearchExtent(15, xb, yb, tid, tidcount)
zoom_depth = 15 # orginal value was 14
# Ok do whatever
filters = tb.Filters(complevel=5, complib='blosc')
atom = tb.Float64Atom()
neogeo_shape = (img_times.size, 64, 64)
times_shape = (img_times.size, )
leaf_ii = 0
for f2l in f2l_mat:
f2l_group = out_tb.create_group(out_tb.root, 'loaded_%d' % int(f2l))
f2l_group.num_feat_loaded = f2l
db_rows, N = load_all_features(dbt, extent, zoom_depth, f2l)
flann, t_build_db = build_flann_db(db_rows)
f2l_group.time_to_build_flann = t_build_db
obs_shape = (img_times.size, db_rows.shape[0], 4)
for n_img in n_img_feat:
n_group = out_tb.create_group(f2l_group, 'feat_%d' % int(n_img))
n_group.img_feat = n_img
neogeo_out = out_tb.create_carray(n_group,
'neogeo_out',
atom=atom,
shape=neogeo_shape,
filters=filters)
obs_out = out_tb.create_carray(n_group,
'obs_out',
atom=atom,
shape=obs_shape,
filters=filters)
# print(neogeo_out)
flann_time = out_tb.create_carray(n_group,
'flann_time',
atom=atom,
shape=times_shape,
filters=filters)
# dbloc = out_tb.create_carray(n_group,'dbloc',atom=atom, shape=dbloc_shape, filters=filters)
for img_num in np.arange(img_times.size):
if f5featmeta.num_feat[img_num] > 0:
generate_obs(img_num, n_img, db_rows, obs_out, neogeo_out,
flann_time, flann)
leaf_values[leaf_ii].value += 1
# if np.mod(leaf_values[leaf_ii].value, 5) == 0:
# prog_tree.cursor.restore()
# prog_tree.draw(test_d)
leaf_ii += 1
out_tb.flush()
out_tb.close()
dbf.close()
f5.close()