Skip to content

Commit

Permalink
Check for no detections
Browse files Browse the repository at this point in the history
  • Loading branch information
@avanetten
avanetten committed Aug 17, 2019
1 parent 2ffd917 commit 38dea74
Showing 1 changed file with 79 additions and 75 deletions.
154 changes: 79 additions & 75 deletions simrdwn/core/simrdwn.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1124,7 +1124,7 @@ def run_test(framework='YOLT2',
# plot

# add geo coords to eall boxes?
if test_add_geo_coords:
if test_add_geo_coords and len(df_tot) > 0:
###########################################
# !!!!! Skip?
# json = None
Expand Down Expand Up @@ -1544,17 +1544,20 @@ def execute(args, train_cmd1, test_cmd_tot, test_cmd_tot2=''):
rotate_boxes=args.rotate_boxes,
test_add_geo_coords=args.test_add_geo_coords)

# save to csv
df_tot.to_csv(args.val_df_path_aug, index=False)
# get number of files
n_files = len(np.unique(df_tot['Loc_Tmp'].values))
# n_files = str(len(test_files_locs_list)
t4 = time.time()
cmd_time_str = '"Length of time to run test for ' \
+ str(n_files) + ' files = ' \
+ str(t4 - t3) + ' seconds\n"'
print(cmd_time_str[1:-1])
os.system('echo ' + cmd_time_str + ' >> ' + args.log_file)
if len(df_tot) == 0:
print("No detections found!")
else:
# save to csv
df_tot.to_csv(args.val_df_path_aug, index=False)
# get number of files
n_files = len(np.unique(df_tot['Loc_Tmp'].values))
# n_files = str(len(test_files_locs_list)
t4 = time.time()
cmd_time_str = '"Length of time to run test for ' \
+ str(n_files) + ' files = ' \
+ str(t4 - t3) + ' seconds\n"'
print(cmd_time_str[1:-1])
os.system('echo ' + cmd_time_str + ' >> ' + args.log_file)

# run again, if desired
if len(args.weight_file2) > 0:
Expand Down Expand Up @@ -1620,69 +1623,70 @@ def execute(args, train_cmd1, test_cmd_tot, test_cmd_tot2=''):
sliced = False
print("test data sliced?", sliced)

# refine for each plot_thresh
for plot_thresh_tmp in args.plot_thresh:
print("Plotting at:", plot_thresh_tmp)
groupby = 'Image_Path'
groupby_cat = 'Category'
df_refine = post_process.refine_df(df_tot,
groupby=groupby,
groupby_cat=groupby_cat,
nms_overlap_thresh=args.nms_overlap_thresh,
plot_thresh=plot_thresh_tmp,
verbose=False)
# make some output plots, if desired
if len(args.building_csv_file) > 0:
building_csv_file_tmp = args.building_csv_file.split('.')[0] \
+ '_plot_thresh_' + str(plot_thresh_tmp).replace('.', 'p') \
+ '.csv'
else:
building_csv_file_tmp = ''
if args.n_test_output_plots > 0:
post_process.plot_refined_df(df_refine, groupby=groupby,
label_map_dict=args.label_map_dict_tot,
outdir=args.results_dir,
plot_thresh=plot_thresh_tmp,
show_labels=bool(
args.show_labels),
alpha_scaling=bool(
args.alpha_scaling),
plot_line_thickness=args.plot_line_thickness,
print_iter=5,
n_plots=args.n_test_output_plots,
building_csv_file=building_csv_file_tmp,
shuffle_ims=bool(
args.shuffle_val_output_plot_ims),
verbose=False)

# geo coords?
if bool(args.test_add_geo_coords):
df_refine, json = add_geo_coords.add_geo_coords_to_df(
df_refine,
create_geojson=bool(args.save_json),
inProj_str='epsg:32737', outProj_str='epsg:3857',
# inProj_str='epsg:4326', outProj_str='epsg:3857',
verbose=False)

# save df_refine
outpath_tmp = os.path.join(args.results_dir,
args.val_prediction_df_refine_tot_root_part +
'_thresh=' + str(plot_thresh_tmp) + '.csv')
# df_refine.to_csv(args.val_prediction_df_refine_tot)
df_refine.to_csv(outpath_tmp)
print("Num objects at thresh:", plot_thresh_tmp, "=",
len(df_refine))
# save json
if bool(args.save_json) and (len(json) > 0):
output_json_path = os.path.join(args.results_dir,
args.val_prediction_df_refine_tot_root_part +
'_thresh=' + str(plot_thresh_tmp) + '.GeoJSON')
json.to_file(output_json_path, driver="GeoJSON")

cmd_time_str = '"Length of time to run refine_test()' + ' ' \
+ str(time.time() - t8) + ' seconds"'
print(cmd_time_str[1:-1])
os.system('echo ' + cmd_time_str + ' >> ' + args.log_file)
# refine for each plot_thresh (if we have detections)
if len(df_tot) > 0:
for plot_thresh_tmp in args.plot_thresh:
print("Plotting at:", plot_thresh_tmp)
groupby = 'Image_Path'
groupby_cat = 'Category'
df_refine = post_process.refine_df(df_tot,
groupby=groupby,
groupby_cat=groupby_cat,
nms_overlap_thresh=args.nms_overlap_thresh,
plot_thresh=plot_thresh_tmp,
verbose=False)
# make some output plots, if desired
if len(args.building_csv_file) > 0:
building_csv_file_tmp = args.building_csv_file.split('.')[0] \
+ '_plot_thresh_' + str(plot_thresh_tmp).replace('.', 'p') \
+ '.csv'
else:
building_csv_file_tmp = ''
if args.n_test_output_plots > 0:
post_process.plot_refined_df(df_refine, groupby=groupby,
label_map_dict=args.label_map_dict_tot,
outdir=args.results_dir,
plot_thresh=plot_thresh_tmp,
show_labels=bool(
args.show_labels),
alpha_scaling=bool(
args.alpha_scaling),
plot_line_thickness=args.plot_line_thickness,
print_iter=5,
n_plots=args.n_test_output_plots,
building_csv_file=building_csv_file_tmp,
shuffle_ims=bool(
args.shuffle_val_output_plot_ims),
verbose=False)

# geo coords?
if bool(args.test_add_geo_coords):
df_refine, json = add_geo_coords.add_geo_coords_to_df(
df_refine,
create_geojson=bool(args.save_json),
inProj_str='epsg:32737', outProj_str='epsg:3857',
# inProj_str='epsg:4326', outProj_str='epsg:3857',
verbose=False)

# save df_refine
outpath_tmp = os.path.join(args.results_dir,
args.val_prediction_df_refine_tot_root_part +
'_thresh=' + str(plot_thresh_tmp) + '.csv')
# df_refine.to_csv(args.val_prediction_df_refine_tot)
df_refine.to_csv(outpath_tmp)
print("Num objects at thresh:", plot_thresh_tmp, "=",
len(df_refine))
# save json
if bool(args.save_json) and (len(json) > 0):
output_json_path = os.path.join(args.results_dir,
args.val_prediction_df_refine_tot_root_part +
'_thresh=' + str(plot_thresh_tmp) + '.GeoJSON')
json.to_file(output_json_path, driver="GeoJSON")

cmd_time_str = '"Length of time to run refine_test()' + ' ' \
+ str(time.time() - t8) + ' seconds"'
print(cmd_time_str[1:-1])
os.system('echo ' + cmd_time_str + ' >> ' + args.log_file)

# remove or zip test_split_dirs to save space
if len(test_split_dir_list) > 0:
Expand Down

0 comments on commit 38dea74

Please sign in to comment.