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

Extracted debug functionality into separate module 

Signed-off-by: Jim Martens <github@2martens.de>
This commit is contained in:
Jim Martens 2019-07-04 17:03:47 +02:00
parent eda3496f04
commit 662c5ec7af
2 changed files with 76 additions and 41 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

46
src/twomartens/masterthesis/cli.py
View File

@ -172,7 +172,8 @@ def _ssd_train(args: argparse.Namespace) -> None:
batch_size = conf.get_property("Parameters.batch_size")
image_size = conf.get_property("Parameters.ssd_image_size")
use_dropout = False if args.network == "ssd" else True
summary_path = conf.get_property("Paths.summaries")
weights_path = conf.get_property("Paths.weights")
pre_trained_weights_file = f"{weights_path}/VGG_coco_SSD_300x300_iter_400000.h5"
weights_path = f"{weights_path}/train/{args.network}/"
@ -213,57 +214,20 @@ def _ssd_train(args: argparse.Namespace) -> None:
del file_names_train, instances_train, file_names_val, instances_val
if args.debug and conf.get_property("Debug.train_images"):
from matplotlib import pyplot
import numpy as np
from PIL import Image
from twomartens.masterthesis.ssd_keras.eval_utils import coco_utils
from twomartens.masterthesis import debug
train_data = next(train_generator)
train_length -= batch_size
train_images = train_data[0]
train_labels = train_data[1]
output_path = f"{summary_path}/train/{args.network}/{args.iteration}"
annotation_file_train = f"{args.coco_path}/annotations/instances_train2014.json"
_, _, _, classes_to_names = coco_utils.get_coco_category_maps(annotation_file_train)
colors = pyplot.cm.hsv(np.linspace(0, 1, 81)).tolist()
summary_path = f"{args.summary_path}/train/{args.network}/{args.iteration}/"
os.makedirs(summary_path, exist_ok=True)
nr_images = len(train_images)
nr_digits = math.ceil(math.log10(nr_images))
for i, train_image in enumerate(train_images):
instances = train_labels[i]
image = Image.fromarray(train_image)
image.save(f"{summary_path}"
f"train_image{str(i).zfill(nr_digits)}.png")
figure = pyplot.figure(figsize=(20, 12))
pyplot.imshow(image)
current_axis = pyplot.gca()
for instance in instances:
xmin = instance[-12] * image_size
ymin = instance[-11] * image_size
xmax = instance[-10] * image_size
ymax = instance[-9] * image_size
class_id = np.argmax(instance[:-12], axis=0)
color = colors[class_id]
label = f"{classes_to_names[class_id]}"
current_axis.add_patch(
pyplot.Rectangle((xmin, ymin), xmax - xmin, ymax - ymin, color=color, fill=False,
linewidth=2))
current_axis.text(xmin, ymin, label, size='x-large', color='white',
bbox={'facecolor': color, 'alpha': 1.0})
pyplot.savefig(f"{args.summary_path}/train/{args.network}/{args.iteration}/bboxes{str(i).zfill(nr_digits)}.png")
pyplot.close(figure)
debug.save_ssd_train_images(train_images, train_labels, output_path)
nr_batches_train = int(math.floor(train_length / batch_size))
nr_batches_val = int(math.floor(val_length / batch_size))
summary_path = conf.get_property("Paths.summaries")
tensorboard_callback = tf.keras.callbacks.TensorBoard(
log_dir=f"{summary_path}/train/{args.network}/{args.iteration}"
)

71
src/twomartens/masterthesis/debug.py Normal file
View File

@ -0,0 +1,71 @@
# -*- coding: utf-8 -*-
# Copyright 2018 Timon Brüning, Inga Kempfert, Anne Kunstmann, Jim Martens,
# Marius Pierenkemper, Yanneck Reiss
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Handle debug functionality.
Functions:
save_ssd_train_images(images, labels, output_path):
saves the first batch of SSD train images with overlaid ground truth bounding boxes
"""
import os
import math
import numpy as np
from matplotlib import pyplot
from PIL import Image
from twomartens.masterthesis import config
from twomartens.masterthesis.ssd_keras.eval_utils import coco_utils
def save_ssd_train_images(images: np.ndarray, labels: np.ndarray, output_path: str) -> None:
annotation_file_train = f"{config.get_property('Paths.coco')}/annotations/instances_train2014.json"
_, _, _, classes_to_names = coco_utils.get_coco_category_maps(annotation_file_train)
colors = pyplot.cm.hsv(np.linspace(0, 1, 81)).tolist()
os.makedirs(output_path, exist_ok=True)
nr_images = len(images)
nr_digits = math.ceil(math.log10(nr_images))
image_size = config.get_property("Parameters.ssd_image_size")
for i, train_image in enumerate(images):
instances = labels[i]
image = Image.fromarray(train_image)
image.save(f"{output_path}"
f"train_image{str(i).zfill(nr_digits)}.png")
figure = pyplot.figure(figsize=(20, 12))
pyplot.imshow(image)
current_axis = pyplot.gca()
for instance in instances:
xmin = instance[-12] * image_size
ymin = instance[-11] * image_size
xmax = instance[-10] * image_size
ymax = instance[-9] * image_size
class_id = np.argmax(instance[:-12], axis=0)
color = colors[class_id]
label = f"{classes_to_names[class_id]}"
current_axis.add_patch(
pyplot.Rectangle((xmin, ymin), xmax - xmin, ymax - ymin, color=color, fill=False,
linewidth=2))
current_axis.text(xmin, ymin, label, size='x-large', color='white',
bbox={'facecolor': color, 'alpha': 1.0})
pyplot.savefig(f"{output_path}/bboxes{str(i).zfill(nr_digits)}.png")
pyplot.close(figure)