From 253a4286741d3382c84d9161bc678b54baf4c4b0 Mon Sep 17 00:00:00 2001
From: LiZeB <963899158@qq.com>
Date: Mon, 13 May 2019 15:59:20 +0800
Subject: [PATCH 1/3] fix the bug which can't use in python3.; Increase the
compatibility between python2. and python3.
---
.idea/MultiNet.iml | 11 ++++++
.idea/inspectionProfiles/Project_Default.xml | 14 ++++++++
.idea/misc.xml | 7 ++++
.idea/modules.xml | 8 +++++
.idea/vcs.xml | 37 ++++++++++++++++++++
5 files changed, 77 insertions(+)
create mode 100644 .idea/MultiNet.iml
create mode 100644 .idea/inspectionProfiles/Project_Default.xml
create mode 100644 .idea/misc.xml
create mode 100644 .idea/modules.xml
create mode 100644 .idea/vcs.xml
diff --git a/.idea/MultiNet.iml b/.idea/MultiNet.iml
new file mode 100644
index 0000000..5c88ce7
--- /dev/null
+++ b/.idea/MultiNet.iml
@@ -0,0 +1,11 @@
+
+
+
+
+
+
+
+
+
+
+
\ No newline at end of file
diff --git a/.idea/inspectionProfiles/Project_Default.xml b/.idea/inspectionProfiles/Project_Default.xml
new file mode 100644
index 0000000..7972536
--- /dev/null
+++ b/.idea/inspectionProfiles/Project_Default.xml
@@ -0,0 +1,14 @@
+
+
+
+
+
+
+
+
\ No newline at end of file
diff --git a/.idea/misc.xml b/.idea/misc.xml
new file mode 100644
index 0000000..3446b3f
--- /dev/null
+++ b/.idea/misc.xml
@@ -0,0 +1,7 @@
+
+
+
+
+
+
+
\ No newline at end of file
diff --git a/.idea/modules.xml b/.idea/modules.xml
new file mode 100644
index 0000000..b9e755e
--- /dev/null
+++ b/.idea/modules.xml
@@ -0,0 +1,8 @@
+
+
+
+
+
+
+
+
\ No newline at end of file
diff --git a/.idea/vcs.xml b/.idea/vcs.xml
new file mode 100644
index 0000000..cf41716
--- /dev/null
+++ b/.idea/vcs.xml
@@ -0,0 +1,37 @@
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
\ No newline at end of file
From 81f6b7a63242fa9a3fd1a9a2a6a106a86ea4f69c Mon Sep 17 00:00:00 2001
From: LiZeB <963899158@qq.com>
Date: Fri, 17 May 2019 10:41:38 +0800
Subject: [PATCH 2/3] =?UTF-8?q?=E5=A2=9E=E5=8A=A0=E4=BA=86=E8=87=AA?=
=?UTF-8?q?=E5=B7=B1=E7=9A=84=E6=B5=8B=E8=AF=95=E6=96=87=E4=BB=B6demo2.py?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
---
demo2.py | 288 +++++++++++++++++++++++++++++++++++++++++++++++
subhypes_test.py | 0
2 files changed, 288 insertions(+)
create mode 100644 demo2.py
create mode 100644 subhypes_test.py
diff --git a/demo2.py b/demo2.py
new file mode 100644
index 0000000..6ad87ac
--- /dev/null
+++ b/demo2.py
@@ -0,0 +1,288 @@
+import os
+import sys
+import cv2
+import logging
+import numpy as np
+from PIL import Image, ImageDraw, ImageFont
+import tensorflow as tf
+sys.path.insert(1, 'incl')
+try:
+ import tensorvision.utils as tv_utils
+ import tensorvision.core as core
+except ImportError:
+ logging.error("Could not import the submodules.")
+ logging.error("Please execute:""'git submodule update --init --recursive'")
+ exit(1)
+sys.path.append("submodules/KittiBox/submodules/")
+try:
+ import utils.train_utils as dec_utils
+except ImportError:
+ logging.error("Could not import submodules/KittiBox/submodules/utils")
+ exit(-1)
+
+flags = tf.app.flags
+FLAGS = flags.FLAGS
+flags.DEFINE_string('logdir', None, 'Path to logdir.')
+flags.DEFINE_string('input', "data/demo/um_000014.png", 'Image to apply KittiSeg.')
+flags.DEFINE_string('output', None, 'Image to apply KittiSeg.')
+logging.basicConfig(format='%(asctime)s %(levelname)s %(message)s', level=logging.INFO, stream=sys.stdout)
+
+def road_draw(image, highway):
+ im = Image.fromarray(image.astype('uint8'))
+ draw = ImageDraw.Draw(im)
+ fnt = ImageFont.truetype('FreeMono/FreeMonoBold.ttf', 40)
+ shape = image.shape
+
+ if highway:
+ draw.text((65, 10), "Highway",
+ font=fnt, fill=(255, 255, 0, 255))
+ draw.ellipse([10, 10, 55, 55], fill=(255, 255, 0, 255),
+ outline=(255, 255, 0, 255))
+ else:
+ draw.text((65, 10), "minor road",
+ font=fnt, fill=(255, 0, 0, 255))
+ draw.ellipse([10, 10, 55, 55], fill=(255, 0, 0, 255),
+ outline=(255, 0, 0, 255))
+
+ return np.array(im).astype('float32')
+
+def load_united_model(logdir):
+ subhypes = {}
+ submodules = {}
+
+ # 加载参数文件，meta_hypes是总的参数文件
+ ### 这一个函数的作用就是从hypes.json这个超参数文件中加载超参数；
+ ### 因为hypes.json中的路径参数是绝对路径，所以要跨平台使用时必须用代码进行修改;
+ ### hypes.json中的参数分为：1.文件路径参数dirs；2.训练时的显示参数logging；
+ ### 3.损失函数的参数：loss_build; 4.模型列表model_list; 5.模型描述文件（三个模型都是用.json文件描述）models;
+ ### 6.优化方法参数solver
+ meta_hypes = tv_utils.load_hypes_from_logdir(logdir, subdir="", base_path='hypes')
+ ###: {'dirs': {'base_path': '/home/lizebin/tensorflow/Detection/MultiNet/hypes',
+ # 'data_dir': 'DATA', 'image_dir': '/home/lizebin/tensorflow/Detection/MultiNet/RUNS/MultiNet_ICCV/images',
+ # 'output_dir': '/home/lizebin/tensorflow/Detection/MultiNet/RUNS/MultiNet_ICCV'},
+ # 'logging': {'display_iter': 50, 'eval_iter': 1000, 'image_iter': 600000, 'save_iter': 5000, 'write_iter': 1000},
+ # 'loss_build': {'recombine': True, 'weighted': False, 'weights': [0.7, 2, 0.7]},
+ # 'model_list': ['segmentation', 'detection', 'road'],
+ # 'models': {'detection': '../submodules/KittiBox/hypes/kittiBox.json',
+ # 'road': '../submodules/KittiClass/hypes/KittiClass_VGG.json',
+ # 'segmentation': '../submodules/KittiSeg/hypes/KittiVGG.json'},
+ # 'path': ['../incl', '../submodules/KittiSeg/incl', '../submodules/KittiBox/incl/', '../submodules/KittiClass/incl/'],
+ # 'selection': {'random': False, 'use_weights': True, 'weights': [1, 0, 0]},
+ # 'solver': {'batch_size': 1, 'max_steps': 100000}}
+
+ #加载模型的各个模块,hypes代表模型的参数文件，modules和submodules[modules]代表模型的组成模块
+ for model in meta_hypes['models']:
+ subhypes[model] = tv_utils.load_hypes_from_logdir(logdir, subdir=model)
+ hypes = subhypes[model]
+ # : {'arch': {'deep_feat': 'pool5'},
+ # 'augment_level': 1, 'batch_size': 5, 'biggest_box_px': 10000, 'clip_norm': 1.0,
+ # 'data': {'eval_cmd': '../submodules/KittiObjective2/./evaluate_object2',
+ # 'label_dir': 'KittiBox/training/label_2', 'train_file': 'KittiBox/train.txt',
+ # 'truncate_data': False, 'val_file': 'KittiBox/val.txt'}, 'detect_truck': False,
+ # 'dirs': {'base_path': '/home/mifs/mttt2/github/MultiNet/submodules/KittiBox/hypes',
+ # 'data_dir': 'DATA', 'image_dir': '/home/lizebin/tensorflow/Detection/MultiNet/RUNS/MultiNet_ICCV/images',
+ # 'output_dir': '/home/lizebin/tensorflow/Detection/MultiNet/RUNS/MultiNet_ICCV'},
+ # 'early_feat_channels': 256, 'focus_size': 1.8, 'grid_height': 12, 'grid_width': 39,
+ # 'image_height': 384, 'image_width': 1248,
+ # 'logging': {'display_iter': 200, 'eval_iter': 2000, 'image_iter': 10000, 'save_iter': 2000,
+ # 'write_iter': 800}, 'model': {'architecture_file': '../encoder/vgg.py',
+ # 'evaluator_file': '../evals/kitti_eval.py', 'input_file': '../inputs/kitti_input.py',
+ # 'objective_file': '../decoder/fastBox.py', 'optimizer_file': '../optimizer/generic_optimizer.py'},
+ # 'num_classes': 2, 'num_inner_channel': 500, 'path': ['../incl'], 'region_size': 32,
+ # 'rezoom_change_loss': 'center', 'rezoom_h_coords': [-0.25, 0, 0.25],
+ # 'rezoom_w_coords': [-0.25, 0, 0.25], 'scale_down': 0.1,
+ # 'solver': {'batch_size': 1, 'epsilon': 1e-05, 'head_weights': [1.0, 0.1],
+ # 'hungarian_iou': 0.35, 'learning_rate': 1e-05, 'learning_rate_step': None,
+ # 'max_steps': 140000, 'opt': 'Adam', 'rnd_seed': 1, 'use_jitter': True, 'weights': ''},
+ # 'tau': 0.35, 'use_mask': True, 'use_rezoom': True, 'wd': 0.0005}
+ hypes['dirs']['output_dir'] = meta_hypes['dirs']['output_dir']
+ hypes['dirs']['image_dir'] = meta_hypes['dirs']['image_dir']
+ submodules[model] = tv_utils.load_modules_from_logdir(logdir, dirname=model, postfix=model)
+
+ ### dict_keys(['input', 'arch', 'objective', 'solver', 'eval'])加载了这5个模块代表一个模型
+ ### submodules[module]代表一个模型的5个模块
+ modules = submodules[model]
+
+ return meta_hypes, subhypes, submodules
+
+def eval_model(logdir, meta_hypes, subhypes, submodules):
+ first_iter = True
+
+ image_pl = tf.placeholder(tf.float32)
+ image = tf.expand_dims(image_pl, 0)
+ image.set_shape([1, 384, 1248, 3])
+
+ decoded_logits = {}
+ for model in meta_hypes['models']:
+ hypes = subhypes[model] ### subhypes[modle]是各个模型的参数文件
+ modules = submodules[model] ### submodules[model]是模型的各个组件
+ optimizer = modules['solver']
+
+ with tf.name_scope('Validation_%s' % model):
+ reuse = {True: False, False: True}[first_iter]
+
+ scope = tf.get_variable_scope()
+
+ with tf.variable_scope(scope, reuse=reuse):
+ ### inference()函数的输出结果是vgg_dict = {'deep_feat': deep_feat,'early_feat': vgg_fcn.conv4_3}
+ ### 这里输出的是一个用来计算交叉熵的tensorflow：softmax_linear: Output tensor with the computed logits.
+ logits = modules['arch'].inference(hypes, image, train=False)
+
+ # """Apply decoder to the logits.
+ #
+ # Computation which decode CNN boxes.
+ # The output can be interpreted as bounding Boxes.
+ # Args:
+ # logits: Logits tensor, output von encoder
+ #
+ # Return:
+ # decoded_logits: values which can be interpreted as bounding boxes
+ # """
+ # decoded_logits[model]是一个字典，是对bounding box的预测。dict_keys(['pred_confs_deltas',
+ # 'pred_boxes_deltas', 'pred_boxes_new', 'pred_boxes', 'pred_logits', 'pred_confidences'])
+ decoded_logits[model] = modules['objective'].decoder(hypes, logits, train=False)
+
+ first_iter = False
+ sess = tf.Session()
+ saver = tf.train.Saver()
+
+ # """
+ # Load the weights of a model stored in saver.
+ #
+ # Parameters
+ # ----------
+ # checkpoint_dir : str
+ # The directory of checkpoints.
+ # sess : tf.Session
+ # A Session to use to restore the parameters.
+ # saver : tf.train.Saver
+ #
+ # Returns
+ # -----------
+ # int
+ # training step of checkpoint
+ # """
+ cur_step = core.load_weights(logdir, sess, saver)
+
+ return meta_hypes, subhypes, submodules, decoded_logits, sess, image_pl
+
+def main():
+ logdir = os.path.join("RUNS", "MultiNet_ICCV")
+ meta_hypes, subhypes, submodules = load_united_model(logdir) ### 从RUN/MultiNet_ICCV下加载模型
+ load_out = eval_model(logdir, meta_hypes, subhypes, submodules)
+
+ # Create list of relevant tensors to evaluate
+ ### meata_hypes是总的参数文件hypes.json;
+ ### subhypes是各个模型的参数文件；
+ ### submoudles是各个模型的组成模块；
+ ### decoded_logits在detection时是bounding box的预测；
+ ### sesss是会话管理环境；
+ ### image_pl是图片输入tensor
+ meta_hypes, subhypes, submodules, decoded_logits, sess, image_pl = load_out
+
+ seg_softmax = decoded_logits['segmentation']['softmax']
+ pred_boxes_new = decoded_logits['detection']['pred_boxes_new']
+ pred_confidences = decoded_logits['detection']['pred_confidences']
+ if len(meta_hypes['model_list']) == 3:
+ road_softmax = decoded_logits['road']['softmax'][0]
+ else:
+ road_softmax = None
+
+ eval_list = [seg_softmax, pred_boxes_new, pred_confidences, road_softmax]
+
+ hypes_road = subhypes['road']
+ image = cv2.imread(FLAGS.input)
+ shape = image.shape
+ image_height = hypes_road['jitter']['image_height']
+ image_width = hypes_road['jitter']['image_width']
+ assert (image_height >= shape[0])
+ assert (image_width >= shape[1])
+ image = cv2.resize(image, (image_height, image_width), interpolation=cv2.INTER_CUBIC)
+
+ ### Run KittiSeg model on image
+ ### eval_list = [seg_softmax, pred_boxes_new, pred_confidences, road_softmax]
+ feed_dict = {image_pl: image}
+ output = sess.run(eval_list, feed_dict=feed_dict)
+ seg_softmax, pred_boxes_new, pred_confidences, road_softmax = output
+
+ # Create Segmentation Overlay
+ # """
+ # Overlay input_image with a hard segmentation result for two classes.
+ # Store the result with the same name as segmentation_image, but with
+ # `-overlay`.
+ #
+ # Parameters
+ # ----------
+ # input_image : numpy.array
+ # An image of shape [width, height, 3].
+ # segmentation : numpy.array
+ # Segmentation of shape [width, height].
+ # color: color for forground class
+ #
+ # Returns
+ # -------
+ # numpy.array
+ # The image overlayed with the segmenation
+ # """
+ shape = image.shape
+ seg_softmax = seg_softmax[:, 1].reshape(shape[0], shape[1])
+ hard = seg_softmax > 0.5
+ overlay_image = tv_utils.fast_overlay(image, hard)
+
+ # Draw Detection Boxes
+ new_img, rects = dec_utils.add_rectangles(
+ subhypes['detection'], [overlay_image], pred_confidences,
+ pred_boxes_new, show_removed=False,
+ use_stitching=True, rnn_len=subhypes['detection']['rnn_len'],
+ min_conf=0.50, tau=subhypes['detection']['tau'])
+
+ # Draw road classification
+ highway = (np.argmax(road_softmax) == 1)
+ new_img = road_draw(new_img, highway)
+
+ ### 在终端上打印出预测的结果：检测到的车辆数；
+ ### Printing some more output information
+ ### removing predictions <= threshold
+ ### Printing coordinates of predicted rects.
+ logging.info("")
+ threshold = 0.5
+ accepted_predictions = []
+ for rect in rects:
+ if rect.score >= threshold:
+ accepted_predictions.append(rect)
+ print('')
+ logging.info("{} Cars detected".format(len(accepted_predictions)))
+ for i, rect in enumerate(accepted_predictions):
+ logging.info("")
+ logging.info("Coordinates of Box {}".format(i))
+ logging.info(" x1: {}".format(rect.x1))
+ logging.info(" x2: {}".format(rect.x2))
+ logging.info(" y1: {}".format(rect.y1))
+ logging.info(" y2: {}".format(rect.y2))
+ logging.info(" Confidence: {}".format(rect.score))
+ if len(meta_hypes['model_list']) == 3:
+ logging.info("Raw Classification Softmax outputs are: {}"
+ .format(output[0][0]))
+
+ # Save output image file
+ if FLAGS.output is None:
+ output_base_name = FLAGS.input
+ out_image_name = output_base_name.split('.')[0] + '_out.png'
+ else:
+ out_image_name = FLAGS.output
+ cv2.imwrite(out_image_name, new_img)
+
+ logging.info("")
+ logging.info("Output image has been saved to: {}".format(os.path.realpath(out_image_name)))
+ logging.info("")
+ logging.warning("Do NOT use this Code to evaluate multiple images.")
+ logging.warning("Demo.py is **very slow** and designed "
+ "to be a tutorial to show how the MultiNet works.")
+ logging.warning("")
+ logging.warning("Please see this comment, if you like to apply demo.py to multiple images see:")
+ logging.warning("https://github.com/MarvinTeichmann/KittiBox/issues/15#issuecomment-301800058")
+
+ exit(0)
+
+if __name__ == '__main__':
+ main()
\ No newline at end of file
diff --git a/subhypes_test.py b/subhypes_test.py
new file mode 100644
index 0000000..e69de29
From ae4c444bbde350b1ebd4e430272fa1b6270f5b2f Mon Sep 17 00:00:00 2001
From: LiZeB <963899158@qq.com>
Date: Fri, 17 May 2019 15:34:49 +0800
Subject: [PATCH 3/3] =?UTF-8?q?=E4=BF=AE=E6=94=B9=E4=BA=86demo.py,?=
=?UTF-8?q?=E5=88=A0=E9=99=A4=E4=BA=86=E4=B8=80=E4=BA=9B=E5=A4=9A=E4=BD=99?=
=?UTF-8?q?=E7=9A=84=E4=BB=A3=E7=A0=81=E3=80=82=E6=8E=A5=E4=B8=8B=E6=9D=A5?=
=?UTF-8?q?=E8=A6=81=E5=AF=B9demo.py=E7=BB=A7=E7=BB=AD=E8=BF=9B=E8=A1=8C?=
=?UTF-8?q?=E4=BF=AE=E6=94=B9=EF=BC=8C=E7=BC=A9=E7=9F=AD=E8=BF=90=E8=A1=8C?=
=?UTF-8?q?=E6=97=B6=E9=97=B4?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
---
demo.py | 187 +++++-------------------------------
demo2.py | 288 -------------------------------------------------------
2 files changed, 25 insertions(+), 450 deletions(-)
delete mode 100644 demo2.py
diff --git a/demo.py b/demo.py
index eca0244..144e640 100644
--- a/demo.py
+++ b/demo.py
@@ -1,12 +1,7 @@
#!/usr/bin/env python
# -*- coding: utf-8 -*-
-#
-# Author: Marvin Teichmann
-
-
-"""
+'''
Detects Cars in an image using KittiSeg.
-
Input: Image
Output: Image (with Cars plotted in Green)
@@ -16,62 +11,36 @@
Usage:
python demo.py --input data/demo.png [--output_image output_image]
[--logdir /path/to/weights] [--gpus 0]
-
-
-"""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
+'''
import json
import logging
import os
import sys
-
import collections
-
-# configure logging
-
-logging.basicConfig(format='%(asctime)s %(levelname)s %(message)s',
- level=logging.INFO,
- stream=sys.stdout)
-
-# https://github.com/tensorflow/tensorflow/issues/2034#issuecomment-220820070
import numpy as np
import scipy as scp
import scipy.misc
import tensorflow as tf
-
import time
-
+import argparse
from PIL import Image, ImageDraw, ImageFont
-
-flags = tf.app.flags
-FLAGS = flags.FLAGS
-
+logging.basicConfig(format='%(asctime)s %(levelname)s %(message)s', level=logging.INFO, stream=sys.stdout)
sys.path.insert(1, 'incl')
-
try:
- # Check whether setup was done correctly
-
import tensorvision.utils as tv_utils
import tensorvision.core as core
except ImportError:
- # You forgot to initialize submodules
logging.error("Could not import the submodules.")
- logging.error("Please execute:"
- "'git submodule update --init --recursive'")
+ logging.error("Please execute:'git submodule update --init --recursive'")
exit(1)
-
-flags.DEFINE_string('logdir', None,
- 'Path to logdir.')
-flags.DEFINE_string('input', None,
- 'Image to apply KittiSeg.')
-flags.DEFINE_string('output', None,
- 'Image to apply KittiSeg.')
-
+parser = argparse.ArgumentParser(argument_default=argparse.SUPPRESS)
+parser.add_argument('--logdir', type=str, default='None', help= "Path to logdir")
+parser.add_argument('--input', type=str, default='data/demo/um_000014.png', help= "Image to apply KittiSeg.")
+parser.add_argument('--output', type=str, default='None', help= "Image to apply KittiSeg.")
+FLAGS = parser.parse_args()
default_run = 'MultiNet_ICCV'
weights_url = ("ftp://mi.eng.cam.ac.uk/"
@@ -98,61 +67,6 @@ def maybe_download_and_extract(runs_dir):
return
-
-def resize_label_image(image, gt_image, image_height, image_width):
- image = scp.misc.imresize(image, size=(image_height, image_width),
- interp='cubic')
- shape = gt_image.shape
- gt_image = scp.misc.imresize(gt_image, size=(image_height, image_width),
- interp='nearest')
-
- return image, gt_image
-
-
-def _output_generator(sess, tensor_list, image_pl, data_file,
- process_image=lambda x: x):
- image_dir = os.path.dirname(data_file)
- with open(data_file) as file:
- for datum in file:
- datum = datum.rstrip()
- image_file = datum.split(" ")[0]
- image_file = os.path.join(image_dir, image_file)
-
- image = scp.misc.imread(image_file)
-
- image = process_image(image)
-
- feed_dict = {image_pl: image}
- start_time = time.time()
- output = sess.run(tensor_list, feed_dict=feed_dict)
- yield image_file, output
-
-
-def eval_runtime(sess, subhypes, image_pl, eval_list, data_file):
- logging.info(' ')
- logging.info('Evaluation complete. Measuring runtime.')
- image_dir = os.path.dirname(data_file)
- with open(data_file) as file:
- for datum in file:
- datum = datum.rstrip()
- image_file = datum.split(" ")[0]
- image_file = os.path.join(image_dir, image_file)
- image = scp.misc.imread(image_file)
- image = process_image(subhypes, image)
- feed = {image_pl: image}
- for i in xrange(100):
- _ = sess.run(eval_list, feed_dict=feed)
- start_time = time.time()
- for i in xrange(100):
- _ = sess.run(eval_list, feed_dict=feed)
- dt = (time.time() - start_time)/100
- logging.info('Joined inference can be conducted at the following rates on'
- ' your machine:')
- logging.info('Speed (msec): %f ', 1000*dt)
- logging.info('Speed (fps): %f ', 1/dt)
- return dt
-
-
def test_constant_input(subhypes):
road_input_conf = subhypes['road']['jitter']
seg_input_conf = subhypes['segmentation']['jitter']
@@ -219,40 +133,17 @@ def road_draw(image, highway):
return np.array(im).astype('float32')
-
-def process_image(subhypes, image):
- hypes = subhypes['road']
- shape = image.shape
- image_height = hypes['jitter']['image_height']
- image_width = hypes['jitter']['image_width']
- assert(image_height >= shape[0])
- assert(image_width >= shape[1])
-
- image = scp.misc.imresize(image, (image_height,
- image_width, 3),
- interp='cubic')
- return image
-
-
def load_united_model(logdir):
subhypes = {}
- subgraph = {}
submodules = {}
- subqueues = {}
-
first_iter = True
-
- meta_hypes = tv_utils.load_hypes_from_logdir(logdir, subdir="",
- base_path='hypes')
+ meta_hypes = tv_utils.load_hypes_from_logdir(logdir, subdir="", base_path='hypes')
for model in meta_hypes['models']:
subhypes[model] = tv_utils.load_hypes_from_logdir(logdir, subdir=model)
hypes = subhypes[model]
hypes['dirs']['output_dir'] = meta_hypes['dirs']['output_dir']
hypes['dirs']['image_dir'] = meta_hypes['dirs']['image_dir']
- submodules[model] = tv_utils.load_modules_from_logdir(logdir,
- dirname=model,
- postfix=model)
-
+ submodules[model] = tv_utils.load_modules_from_logdir(logdir, dirname=model, postfix=model)
modules = submodules[model]
image_pl = tf.placeholder(tf.float32)
@@ -266,14 +157,10 @@ def load_united_model(logdir):
with tf.name_scope('Validation_%s' % model):
reuse = {True: False, False: True}[first_iter]
-
scope = tf.get_variable_scope()
-
with tf.variable_scope(scope, reuse=reuse):
logits = modules['arch'].inference(hypes, image, train=False)
-
- decoded_logits[model] = modules['objective'].decoder(hypes, logits,
- train=False)
+ decoded_logits[model] = modules['objective'].decoder(hypes, logits, train=False)
first_iter = False
sess = tf.Session()
@@ -283,10 +170,10 @@ def load_united_model(logdir):
return meta_hypes, subhypes, submodules, decoded_logits, sess, image_pl
-def main(_):
+def main():
tv_utils.set_gpus_to_use()
- if FLAGS.input is None:
+ if FLAGS.input == 'None':
logging.error("No input was given.")
logging.info(
"Usage: python demo.py --input data/test.png "
@@ -294,13 +181,8 @@ def main(_):
"[--gpus GPUs_to_use] ")
exit(1)
- if FLAGS.logdir is None:
- # Download and use weights from the MultiNet Paper
- if 'TV_DIR_RUNS' in os.environ:
- runs_dir = os.path.join(os.environ['TV_DIR_RUNS'],
- 'MultiNet')
- else:
- runs_dir = 'RUNS'
+ if FLAGS.logdir == 'None':
+ runs_dir = 'RUNS'
maybe_download_and_extract(runs_dir)
logdir = os.path.join(runs_dir, default_run)
else:
@@ -330,6 +212,7 @@ def main(_):
test_segmentation_input(subhypes)
# Load and reseize Image
+
image_file = FLAGS.input
image = scp.misc.imread(image_file)
@@ -339,17 +222,12 @@ def main(_):
image_width = hypes_road['jitter']['image_width']
assert(image_height >= shape[0])
assert(image_width >= shape[1])
-
- image = scp.misc.imresize(image, (image_height,
- image_width, 3),
- interp='cubic')
-
- import utils.train_utils as dec_utils
+ image = scp.misc.imresize(image, (image_height, image_width, 3), interp='cubic')
# Run KittiSeg model on image
feed_dict = {image_pl: image}
output = sess.run(eval_list, feed_dict=feed_dict)
-
+ sess.close()
seg_softmax, pred_boxes_new, pred_confidences, road_softmax = output
# Create Segmentation Overlay
@@ -359,6 +237,7 @@ def main(_):
overlay_image = tv_utils.fast_overlay(image, hard)
# Draw Detection Boxes
+ import utils.train_utils as dec_utils
new_img, rects = dec_utils.add_rectangles(
subhypes['detection'], [overlay_image], pred_confidences,
pred_boxes_new, show_removed=False,
@@ -370,7 +249,6 @@ def main(_):
new_img = road_draw(new_img, highway)
logging.info("")
-
# Printing some more output information
threshold = 0.5
accepted_predictions = []
@@ -393,34 +271,19 @@ def main(_):
logging.info(" Confidence: {}".format(rect.score))
if len(meta_hypes['model_list']) == 3:
- logging.info("Raw Classification Softmax outputs are: {}"
- .format(output[0][0]))
+ logging.info("Raw Classification Softmax outputs are: {}".format(output[0][0]))
# Save output image file
- if FLAGS.output is None:
+ if FLAGS.output == 'None':
output_base_name = FLAGS.input
out_image_name = output_base_name.split('.')[0] + '_out.png'
else:
out_image_name = FLAGS.output
-
scp.misc.imsave(out_image_name, new_img)
logging.info("")
- logging.info("Output image has been saved to: {}".format(
- os.path.realpath(out_image_name)))
-
- logging.info("")
- logging.warning("Do NOT use this Code to evaluate multiple images.")
-
- logging.warning("Demo.py is **very slow** and designed "
- "to be a tutorial to show how the MultiNet works.")
- logging.warning("")
- logging.warning("Please see this comment, if you like to apply demo.py to"
- " multiple images see:")
- logging.warning("https://github.com/MarvinTeichmann/KittiBox/"
- "issues/15#issuecomment-301800058")
-
+ logging.info("Output image has been saved to: {}".format(os.path.realpath(out_image_name)))
exit(0)
if __name__ == '__main__':
- tf.app.run()
+ main()
diff --git a/demo2.py b/demo2.py
deleted file mode 100644
index 6ad87ac..0000000
--- a/demo2.py
+++ /dev/null
@@ -1,288 +0,0 @@
-import os
-import sys
-import cv2
-import logging
-import numpy as np
-from PIL import Image, ImageDraw, ImageFont
-import tensorflow as tf
-sys.path.insert(1, 'incl')
-try:
- import tensorvision.utils as tv_utils
- import tensorvision.core as core
-except ImportError:
- logging.error("Could not import the submodules.")
- logging.error("Please execute:""'git submodule update --init --recursive'")
- exit(1)
-sys.path.append("submodules/KittiBox/submodules/")
-try:
- import utils.train_utils as dec_utils
-except ImportError:
- logging.error("Could not import submodules/KittiBox/submodules/utils")
- exit(-1)
-
-flags = tf.app.flags
-FLAGS = flags.FLAGS
-flags.DEFINE_string('logdir', None, 'Path to logdir.')
-flags.DEFINE_string('input', "data/demo/um_000014.png", 'Image to apply KittiSeg.')
-flags.DEFINE_string('output', None, 'Image to apply KittiSeg.')
-logging.basicConfig(format='%(asctime)s %(levelname)s %(message)s', level=logging.INFO, stream=sys.stdout)
-
-def road_draw(image, highway):
- im = Image.fromarray(image.astype('uint8'))
- draw = ImageDraw.Draw(im)
- fnt = ImageFont.truetype('FreeMono/FreeMonoBold.ttf', 40)
- shape = image.shape
-
- if highway:
- draw.text((65, 10), "Highway",
- font=fnt, fill=(255, 255, 0, 255))
- draw.ellipse([10, 10, 55, 55], fill=(255, 255, 0, 255),
- outline=(255, 255, 0, 255))
- else:
- draw.text((65, 10), "minor road",
- font=fnt, fill=(255, 0, 0, 255))
- draw.ellipse([10, 10, 55, 55], fill=(255, 0, 0, 255),
- outline=(255, 0, 0, 255))
-
- return np.array(im).astype('float32')
-
-def load_united_model(logdir):
- subhypes = {}
- submodules = {}
-
- # 加载参数文件，meta_hypes是总的参数文件
- ### 这一个函数的作用就是从hypes.json这个超参数文件中加载超参数；
- ### 因为hypes.json中的路径参数是绝对路径，所以要跨平台使用时必须用代码进行修改;
- ### hypes.json中的参数分为：1.文件路径参数dirs；2.训练时的显示参数logging；
- ### 3.损失函数的参数：loss_build; 4.模型列表model_list; 5.模型描述文件（三个模型都是用.json文件描述）models;
- ### 6.优化方法参数solver
- meta_hypes = tv_utils.load_hypes_from_logdir(logdir, subdir="", base_path='hypes')
- ###: {'dirs': {'base_path': '/home/lizebin/tensorflow/Detection/MultiNet/hypes',
- # 'data_dir': 'DATA', 'image_dir': '/home/lizebin/tensorflow/Detection/MultiNet/RUNS/MultiNet_ICCV/images',
- # 'output_dir': '/home/lizebin/tensorflow/Detection/MultiNet/RUNS/MultiNet_ICCV'},
- # 'logging': {'display_iter': 50, 'eval_iter': 1000, 'image_iter': 600000, 'save_iter': 5000, 'write_iter': 1000},
- # 'loss_build': {'recombine': True, 'weighted': False, 'weights': [0.7, 2, 0.7]},
- # 'model_list': ['segmentation', 'detection', 'road'],
- # 'models': {'detection': '../submodules/KittiBox/hypes/kittiBox.json',
- # 'road': '../submodules/KittiClass/hypes/KittiClass_VGG.json',
- # 'segmentation': '../submodules/KittiSeg/hypes/KittiVGG.json'},
- # 'path': ['../incl', '../submodules/KittiSeg/incl', '../submodules/KittiBox/incl/', '../submodules/KittiClass/incl/'],
- # 'selection': {'random': False, 'use_weights': True, 'weights': [1, 0, 0]},
- # 'solver': {'batch_size': 1, 'max_steps': 100000}}
-
- #加载模型的各个模块,hypes代表模型的参数文件，modules和submodules[modules]代表模型的组成模块
- for model in meta_hypes['models']:
- subhypes[model] = tv_utils.load_hypes_from_logdir(logdir, subdir=model)
- hypes = subhypes[model]
- # : {'arch': {'deep_feat': 'pool5'},
- # 'augment_level': 1, 'batch_size': 5, 'biggest_box_px': 10000, 'clip_norm': 1.0,
- # 'data': {'eval_cmd': '../submodules/KittiObjective2/./evaluate_object2',
- # 'label_dir': 'KittiBox/training/label_2', 'train_file': 'KittiBox/train.txt',
- # 'truncate_data': False, 'val_file': 'KittiBox/val.txt'}, 'detect_truck': False,
- # 'dirs': {'base_path': '/home/mifs/mttt2/github/MultiNet/submodules/KittiBox/hypes',
- # 'data_dir': 'DATA', 'image_dir': '/home/lizebin/tensorflow/Detection/MultiNet/RUNS/MultiNet_ICCV/images',
- # 'output_dir': '/home/lizebin/tensorflow/Detection/MultiNet/RUNS/MultiNet_ICCV'},
- # 'early_feat_channels': 256, 'focus_size': 1.8, 'grid_height': 12, 'grid_width': 39,
- # 'image_height': 384, 'image_width': 1248,
- # 'logging': {'display_iter': 200, 'eval_iter': 2000, 'image_iter': 10000, 'save_iter': 2000,
- # 'write_iter': 800}, 'model': {'architecture_file': '../encoder/vgg.py',
- # 'evaluator_file': '../evals/kitti_eval.py', 'input_file': '../inputs/kitti_input.py',
- # 'objective_file': '../decoder/fastBox.py', 'optimizer_file': '../optimizer/generic_optimizer.py'},
- # 'num_classes': 2, 'num_inner_channel': 500, 'path': ['../incl'], 'region_size': 32,
- # 'rezoom_change_loss': 'center', 'rezoom_h_coords': [-0.25, 0, 0.25],
- # 'rezoom_w_coords': [-0.25, 0, 0.25], 'scale_down': 0.1,
- # 'solver': {'batch_size': 1, 'epsilon': 1e-05, 'head_weights': [1.0, 0.1],
- # 'hungarian_iou': 0.35, 'learning_rate': 1e-05, 'learning_rate_step': None,
- # 'max_steps': 140000, 'opt': 'Adam', 'rnd_seed': 1, 'use_jitter': True, 'weights': ''},
- # 'tau': 0.35, 'use_mask': True, 'use_rezoom': True, 'wd': 0.0005}
- hypes['dirs']['output_dir'] = meta_hypes['dirs']['output_dir']
- hypes['dirs']['image_dir'] = meta_hypes['dirs']['image_dir']
- submodules[model] = tv_utils.load_modules_from_logdir(logdir, dirname=model, postfix=model)
-
- ### dict_keys(['input', 'arch', 'objective', 'solver', 'eval'])加载了这5个模块代表一个模型
- ### submodules[module]代表一个模型的5个模块
- modules = submodules[model]
-
- return meta_hypes, subhypes, submodules
-
-def eval_model(logdir, meta_hypes, subhypes, submodules):
- first_iter = True
-
- image_pl = tf.placeholder(tf.float32)
- image = tf.expand_dims(image_pl, 0)
- image.set_shape([1, 384, 1248, 3])
-
- decoded_logits = {}
- for model in meta_hypes['models']:
- hypes = subhypes[model] ### subhypes[modle]是各个模型的参数文件
- modules = submodules[model] ### submodules[model]是模型的各个组件
- optimizer = modules['solver']
-
- with tf.name_scope('Validation_%s' % model):
- reuse = {True: False, False: True}[first_iter]
-
- scope = tf.get_variable_scope()
-
- with tf.variable_scope(scope, reuse=reuse):
- ### inference()函数的输出结果是vgg_dict = {'deep_feat': deep_feat,'early_feat': vgg_fcn.conv4_3}
- ### 这里输出的是一个用来计算交叉熵的tensorflow：softmax_linear: Output tensor with the computed logits.
- logits = modules['arch'].inference(hypes, image, train=False)
-
- # """Apply decoder to the logits.
- #
- # Computation which decode CNN boxes.
- # The output can be interpreted as bounding Boxes.
- # Args:
- # logits: Logits tensor, output von encoder
- #
- # Return:
- # decoded_logits: values which can be interpreted as bounding boxes
- # """
- # decoded_logits[model]是一个字典，是对bounding box的预测。dict_keys(['pred_confs_deltas',
- # 'pred_boxes_deltas', 'pred_boxes_new', 'pred_boxes', 'pred_logits', 'pred_confidences'])
- decoded_logits[model] = modules['objective'].decoder(hypes, logits, train=False)
-
- first_iter = False
- sess = tf.Session()
- saver = tf.train.Saver()
-
- # """
- # Load the weights of a model stored in saver.
- #
- # Parameters
- # ----------
- # checkpoint_dir : str
- # The directory of checkpoints.
- # sess : tf.Session
- # A Session to use to restore the parameters.
- # saver : tf.train.Saver
- #
- # Returns
- # -----------
- # int
- # training step of checkpoint
- # """
- cur_step = core.load_weights(logdir, sess, saver)
-
- return meta_hypes, subhypes, submodules, decoded_logits, sess, image_pl
-
-def main():
- logdir = os.path.join("RUNS", "MultiNet_ICCV")
- meta_hypes, subhypes, submodules = load_united_model(logdir) ### 从RUN/MultiNet_ICCV下加载模型
- load_out = eval_model(logdir, meta_hypes, subhypes, submodules)
-
- # Create list of relevant tensors to evaluate
- ### meata_hypes是总的参数文件hypes.json;
- ### subhypes是各个模型的参数文件；
- ### submoudles是各个模型的组成模块；
- ### decoded_logits在detection时是bounding box的预测；
- ### sesss是会话管理环境；
- ### image_pl是图片输入tensor
- meta_hypes, subhypes, submodules, decoded_logits, sess, image_pl = load_out
-
- seg_softmax = decoded_logits['segmentation']['softmax']
- pred_boxes_new = decoded_logits['detection']['pred_boxes_new']
- pred_confidences = decoded_logits['detection']['pred_confidences']
- if len(meta_hypes['model_list']) == 3:
- road_softmax = decoded_logits['road']['softmax'][0]
- else:
- road_softmax = None
-
- eval_list = [seg_softmax, pred_boxes_new, pred_confidences, road_softmax]
-
- hypes_road = subhypes['road']
- image = cv2.imread(FLAGS.input)
- shape = image.shape
- image_height = hypes_road['jitter']['image_height']
- image_width = hypes_road['jitter']['image_width']
- assert (image_height >= shape[0])
- assert (image_width >= shape[1])
- image = cv2.resize(image, (image_height, image_width), interpolation=cv2.INTER_CUBIC)
-
- ### Run KittiSeg model on image
- ### eval_list = [seg_softmax, pred_boxes_new, pred_confidences, road_softmax]
- feed_dict = {image_pl: image}
- output = sess.run(eval_list, feed_dict=feed_dict)
- seg_softmax, pred_boxes_new, pred_confidences, road_softmax = output
-
- # Create Segmentation Overlay
- # """
- # Overlay input_image with a hard segmentation result for two classes.
- # Store the result with the same name as segmentation_image, but with
- # `-overlay`.
- #
- # Parameters
- # ----------
- # input_image : numpy.array
- # An image of shape [width, height, 3].
- # segmentation : numpy.array
- # Segmentation of shape [width, height].
- # color: color for forground class
- #
- # Returns
- # -------
- # numpy.array
- # The image overlayed with the segmenation
- # """
- shape = image.shape
- seg_softmax = seg_softmax[:, 1].reshape(shape[0], shape[1])
- hard = seg_softmax > 0.5
- overlay_image = tv_utils.fast_overlay(image, hard)
-
- # Draw Detection Boxes
- new_img, rects = dec_utils.add_rectangles(
- subhypes['detection'], [overlay_image], pred_confidences,
- pred_boxes_new, show_removed=False,
- use_stitching=True, rnn_len=subhypes['detection']['rnn_len'],
- min_conf=0.50, tau=subhypes['detection']['tau'])
-
- # Draw road classification
- highway = (np.argmax(road_softmax) == 1)
- new_img = road_draw(new_img, highway)
-
- ### 在终端上打印出预测的结果：检测到的车辆数；
- ### Printing some more output information
- ### removing predictions <= threshold
- ### Printing coordinates of predicted rects.
- logging.info("")
- threshold = 0.5
- accepted_predictions = []
- for rect in rects:
- if rect.score >= threshold:
- accepted_predictions.append(rect)
- print('')
- logging.info("{} Cars detected".format(len(accepted_predictions)))
- for i, rect in enumerate(accepted_predictions):
- logging.info("")
- logging.info("Coordinates of Box {}".format(i))
- logging.info(" x1: {}".format(rect.x1))
- logging.info(" x2: {}".format(rect.x2))
- logging.info(" y1: {}".format(rect.y1))
- logging.info(" y2: {}".format(rect.y2))
- logging.info(" Confidence: {}".format(rect.score))
- if len(meta_hypes['model_list']) == 3:
- logging.info("Raw Classification Softmax outputs are: {}"
- .format(output[0][0]))
-
- # Save output image file
- if FLAGS.output is None:
- output_base_name = FLAGS.input
- out_image_name = output_base_name.split('.')[0] + '_out.png'
- else:
- out_image_name = FLAGS.output
- cv2.imwrite(out_image_name, new_img)
-
- logging.info("")
- logging.info("Output image has been saved to: {}".format(os.path.realpath(out_image_name)))
- logging.info("")
- logging.warning("Do NOT use this Code to evaluate multiple images.")
- logging.warning("Demo.py is **very slow** and designed "
- "to be a tutorial to show how the MultiNet works.")
- logging.warning("")
- logging.warning("Please see this comment, if you like to apply demo.py to multiple images see:")
- logging.warning("https://github.com/MarvinTeichmann/KittiBox/issues/15#issuecomment-301800058")
-
- exit(0)
-
-if __name__ == '__main__':
- main()
\ No newline at end of file