Nesne Algılama

Bu İşbirliği, nesne algılama gerçekleştirmek için eğitilmiş bir TF-Hub modülünün kullanımını gösterir.

Kurmak

İthalat ve fonksiyon tanımları

# For running inference on the TF-Hub module.
import tensorflow as tf

import tensorflow_hub as hub

# For downloading the image.
import matplotlib.pyplot as plt
import tempfile
from six.moves.urllib.request import urlopen
from six import BytesIO

# For drawing onto the image.
import numpy as np
from PIL import Image
from PIL import ImageColor
from PIL import ImageDraw
from PIL import ImageFont
from PIL import ImageOps

# For measuring the inference time.
import time

# Print Tensorflow version
print(tf.__version__)

# Check available GPU devices.
print("The following GPU devices are available: %s" % tf.test.gpu_device_name())
2.7.0
The following GPU devices are available: /device:GPU:0

Örnek kullanım

Görüntüleri indirmek ve görselleştirme için yardımcı işlevler.

Görselleştirme kodu uyarlanmıştır TF nesne algılama API basit istenen işlevlerini.

def display_image(image):
  fig
= plt.figure(figsize=(20, 15))
  plt
.grid(False)
  plt
.imshow(image)


def download_and_resize_image(url, new_width=256, new_height=256,
                              display
=False):
  _
, filename = tempfile.mkstemp(suffix=".jpg")
  response
= urlopen(url)
  image_data
= response.read()
  image_data
= BytesIO(image_data)
  pil_image
= Image.open(image_data)
  pil_image
= ImageOps.fit(pil_image, (new_width, new_height), Image.ANTIALIAS)
  pil_image_rgb
= pil_image.convert("RGB")
  pil_image_rgb
.save(filename, format="JPEG", quality=90)
 
print("Image downloaded to %s." % filename)
 
if display:
    display_image
(pil_image)
 
return filename


def draw_bounding_box_on_image(image,
                               ymin
,
                               xmin
,
                               ymax
,
                               xmax
,
                               color
,
                               font
,
                               thickness
=4,
                               display_str_list
=()):
 
"""Adds a bounding box to an image."""
  draw
= ImageDraw.Draw(image)
  im_width
, im_height = image.size
 
(left, right, top, bottom) = (xmin * im_width, xmax * im_width,
                                ymin
* im_height, ymax * im_height)
  draw
.line([(left, top), (left, bottom), (right, bottom), (right, top),
             
(left, top)],
            width
=thickness,
            fill
=color)

 
# If the total height of the display strings added to the top of the bounding
 
# box exceeds the top of the image, stack the strings below the bounding box
 
# instead of above.
  display_str_heights
= [font.getsize(ds)[1] for ds in display_str_list]
 
# Each display_str has a top and bottom margin of 0.05x.
  total_display_str_height
= (1 + 2 * 0.05) * sum(display_str_heights)

 
if top > total_display_str_height:
    text_bottom
= top
 
else:
    text_bottom
= top + total_display_str_height
 
# Reverse list and print from bottom to top.
 
for display_str in display_str_list[::-1]:
    text_width
, text_height = font.getsize(display_str)
    margin
= np.ceil(0.05 * text_height)
    draw
.rectangle([(left, text_bottom - text_height - 2 * margin),
                   
(left + text_width, text_bottom)],
                   fill
=color)
    draw
.text((left + margin, text_bottom - text_height - margin),
              display_str
,
              fill
="black",
              font
=font)
    text_bottom
-= text_height - 2 * margin


def draw_boxes(image, boxes, class_names, scores, max_boxes=10, min_score=0.1):
 
"""Overlay labeled boxes on an image with formatted scores and label names."""
  colors
= list(ImageColor.colormap.values())

 
try:
    font
= ImageFont.truetype("/usr/share/fonts/truetype/liberation/LiberationSansNarrow-Regular.ttf",
                             
25)
 
except IOError:
   
print("Font not found, using default font.")
    font
= ImageFont.load_default()

 
for i in range(min(boxes.shape[0], max_boxes)):
   
if scores[i] >= min_score:
      ymin
, xmin, ymax, xmax = tuple(boxes[i])
      display_str
= "{}: {}%".format(class_names[i].decode("ascii"),
                                     
int(100 * scores[i]))
      color
= colors[hash(class_names[i]) % len(colors)]
      image_pil
= Image.fromarray(np.uint8(image)).convert("RGB")
      draw_bounding_box_on_image
(
          image_pil
,
          ymin
,
          xmin
,
          ymax
,
          xmax
,
          color
,
          font
,
          display_str_list
=[display_str])
      np
.copyto(image, np.array(image_pil))
 
return image

Modülü uygula

Open Images v4'ten genel bir resim yükleyin ve yerel olarak kaydedin ve görüntüleyin.

# By Heiko Gorski, Source: https://commons.wikimedia.org/wiki/File:Naxos_Taverna.jpg
image_url
= "https://upload.wikimedia.org/wikipedia/commons/6/60/Naxos_Taverna.jpg"
downloaded_image_path
= download_and_resize_image(image_url, 1280, 856, True)
Image downloaded to /tmp/tmpu_02gvdt.jpg.

png

Bir nesne algılama modülü seçin ve indirilen görüntüye uygulayın. Modüller:

  • FasterRCNN + InceptionResNet V2: yüksek doğruluk,
  • ssd + MobileNet V2: hızlı, küçük ve.
module_handle = "https://tfhub.dev/google/faster_rcnn/openimages_v4/inception_resnet_v2/1"

detector
= hub.load(module_handle).signatures['default']
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
INFO:tensorflow:Saver not created because there are no variables in the graph to restore
def load_img(path):
  img
= tf.io.read_file(path)
  img
= tf.image.decode_jpeg(img, channels=3)
 
return img
def run_detector(detector, path):
  img
= load_img(path)

  converted_img  
= tf.image.convert_image_dtype(img, tf.float32)[tf.newaxis, ...]
  start_time
= time.time()
  result
= detector(converted_img)
  end_time
= time.time()

  result
= {key:value.numpy() for key,value in result.items()}

 
print("Found %d objects." % len(result["detection_scores"]))
 
print("Inference time: ", end_time-start_time)

  image_with_boxes
= draw_boxes(
      img
.numpy(), result["detection_boxes"],
      result
["detection_class_entities"], result["detection_scores"])

  display_image
(image_with_boxes)
run_detector(detector, downloaded_image_path)
Found 100 objects.
Inference time:  37.78577899932861

png

Daha fazla resim

Zaman takibi ile bazı ek görüntüler üzerinde çıkarım yapın.

image_urls = [
 
# Source: https://commons.wikimedia.org/wiki/File:The_Coleoptera_of_the_British_islands_(Plate_125)_(8592917784).jpg
 
"https://upload.wikimedia.org/wikipedia/commons/1/1b/The_Coleoptera_of_the_British_islands_%28Plate_125%29_%288592917784%29.jpg",
 
# By Américo Toledano, Source: https://commons.wikimedia.org/wiki/File:Biblioteca_Maim%C3%B3nides,_Campus_Universitario_de_Rabanales_007.jpg
 
"https://upload.wikimedia.org/wikipedia/commons/thumb/0/0d/Biblioteca_Maim%C3%B3nides%2C_Campus_Universitario_de_Rabanales_007.jpg/1024px-Biblioteca_Maim%C3%B3nides%2C_Campus_Universitario_de_Rabanales_007.jpg",
 
# Source: https://commons.wikimedia.org/wiki/File:The_smaller_British_birds_(8053836633).jpg
 
"https://upload.wikimedia.org/wikipedia/commons/0/09/The_smaller_British_birds_%288053836633%29.jpg",
 
]

def detect_img(image_url):
  start_time
= time.time()
  image_path
= download_and_resize_image(image_url, 640, 480)
  run_detector
(detector, image_path)
  end_time
= time.time()
 
print("Inference time:",end_time-start_time)
detect_img(image_urls[0])
Image downloaded to /tmp/tmpuxkybwg_.jpg.
Found 100 objects.
Inference time:  1.385962724685669
Inference time: 1.8049812316894531

png

detect_img(image_urls[1])
Image downloaded to /tmp/tmp3wrs8a5l.jpg.
Found 100 objects.
Inference time:  0.8379817008972168
Inference time: 1.2247464656829834

png

detect_img(image_urls[2])
Image downloaded to /tmp/tmpu5bhhlnw.jpg.
Found 100 objects.
Inference time:  0.8334732055664062
Inference time: 1.394953966140747

png