2martens/masterthesis
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Added application of inverse transforms and removal of dummy predictions

Browse Source 
Signed-off-by: Jim Martens <github@2martens.de>
This commit is contained in:
Jim Martens
2019-05-15 14:30:53 +02:00
parent dcee6643fe
commit 9f5bea9c88
1 changed files with 36 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

44
src/twomartens/masterthesis/ssd.py
View File

@ -35,6 +35,7 @@ Functions:
train(...): trains the SSD/DropoutSSD on a given data set train(...): trains the SSD/DropoutSSD on a given data set
""" """
import os import os
import pickle
import time import time
from typing import Dict from typing import Dict
from typing import Optional from typing import Optional
@ -191,28 +192,53 @@ def _predict_one_epoch(dataset: tf.data.Dataset,
if use_dropout: if use_dropout:
for _ in range(forward_passes_per_image): for _ in range(forward_passes_per_image):
result = np.array(ssd(inputs)) result = np.array(ssd(inputs))
result_filtered = []
# iterate over result of images
for i in range(result.shape[0]):
# apply inverse transformations to predicted bounding box coordinates
# filter out dummy all-zero results
x_reverse = labels[i, 0, 5]
y_reverse = labels[i, 0, 6]
filtered = result[i][result[i, :, 0] != 0]
filtered[:, 2] *= x_reverse
filtered[:, 4] *= x_reverse
filtered[:, 3] *= y_reverse
filtered[:, 5] *= y_reverse
result_filtered.append(filtered)
result = result_filtered
decoded_predictions_batch.append(result) decoded_predictions_batch.append(result)
del result del result
else: else:
result = np.array(ssd(inputs)) result = np.array(ssd(inputs))
result_filtered = []
# iterate over result of images
for i in range(result.shape[0]):
# apply inverse transformations to predicted bounding box coordinates
# filter out dummy all-zero results
x_reverse = labels[i, 0, 5]
y_reverse = labels[i, 0, 6]
filtered = result[i][result[i, :, 0] != 0]
filtered[:, 2] *= x_reverse
filtered[:, 4] *= x_reverse
filtered[:, 3] *= y_reverse
filtered[:, 5] *= y_reverse
result_filtered.append(filtered)
result = result_filtered
decoded_predictions_batch.append(result) decoded_predictions_batch.append(result)
del result del result
# save predictions batch-wise to prevent memory problems # save predictions batch-wise to prevent memory problems
if nr_digits is not None: if nr_digits is not None:
counter_str = str(counter).zfill(nr_digits) counter_str = str(counter).zfill(nr_digits)
filename = f"{output_file}-{counter_str}.npy" filename = f"{output_file}-{counter_str}.bin"
label_filename = f"{label_output_file}-{counter_str}.npy" label_filename = f"{label_output_file}-{counter_str}.bin"
else: else:
filename = f"{output_file}-{counter:d}.npy" filename = f"{output_file}-{counter:d}.bin"
label_filename = f"{label_output_file}-{counter:d}.npy" label_filename = f"{label_output_file}-{counter:d}.bin"
with open(filename, 'wb') as file, open(label_filename, 'wb') as label_file: with open(filename, 'wb') as file, open(label_filename, 'wb') as label_file:
decoded_predictions_batch_np = np.array(decoded_predictions_batch) pickle.dump(decoded_predictions_batch, file)
del decoded_predictions_batch pickle.dump(labels, label_file)
np.save(file, decoded_predictions_batch_np, allow_pickle=False, fix_imports=False)
del decoded_predictions_batch_np
np.save(label_file, labels, allow_pickle=False, fix_imports=False)
counter += 1 counter += 1