Module ambianic.pipeline.ai.tf_detect
Tensorflow image detection wrapper.
Expand source code
"""Tensorflow image detection wrapper."""
import logging
import re
import time
import numpy as np
from ambianic.pipeline import PipeElement
# from importlib import import_module
from PIL import ImageOps
from .inference import TFInferenceEngine
log = logging.getLogger(__name__)
class TFDetectionModel(PipeElement):
"""Applies Tensorflow image detection."""
def __init__(
self,
model=None,
labels=None,
label_filter=None,
confidence_threshold=0.6,
top_k=3,
**kwargs,
):
"""Initialize detector with config parameters.
:Parameters:
----------
model: ai_models/mobilenet_ssd_v2_face.tflite
labels: ai_models/coco_labels.txt
confidence_threshold: 0.6
top_k: 3
"""
# log.warning('TFImageDetection __init__ invoked')
super().__init__(**kwargs)
self._tfengine = TFInferenceEngine(
model=model,
labels=labels,
confidence_threshold=confidence_threshold,
top_k=top_k,
)
self._labels = self.load_labels(self._tfengine.labels_path)
self._label_filter = label_filter
self.last_time = time.monotonic()
def load_labels(self, label_path=None):
"""Load label mapping from integer code to text.
:Parameters:
----------
label_path : string
Path to label mapping file.
:Returns:
-------
dict
{label_code, label_text}
"""
assert label_path
p = re.compile(r"\s*(\d+)(.+)")
with open(label_path, encoding="utf-8") as f:
lines = (p.match(line).groups() for line in f.readlines())
return {int(num): text.strip() for num, text in lines}
@staticmethod
def thumbnail(image=None, desired_size=None):
"""Resizes original image as close as possible to desired size.
Preserves aspect ratio of original image.
Does not modify the original image.
:Parameters:
----------
image : PIL.Image
Input Image for AI model detection.
desired_size : (width, height)
Size expected by the AI model.
:Returns:
-------
PIL.Image
Resized image fitting for the AI model input tensor.
"""
assert image
assert desired_size
log.debug("input image size = %r", image.size)
thumb = image.copy()
w, h = desired_size
try:
# convert from numpy to native Python int type
# that PIL expects
if isinstance(w, np.generic):
w = w.item()
w = int(w)
h = h.item()
h = int(h)
thumb.thumbnail((w, h))
except Exception as e:
msg = (
f"Exception in "
f"PIL.image.thumbnail(desired_size={desired_size}):"
f"type(width)={type(w)}, type(height)={type(h)}"
f"\n{e}"
)
log.exception(msg)
raise RuntimeError(msg)
log.debug("thmubnail image size = %r", thumb.size)
return thumb
@staticmethod
def resize(image=None, desired_size=None):
"""Pad original image to exact size expected by input tensor.
Preserve aspect ratio to avoid confusing the AI model with
unnatural distortions. Pad the resulting image
with solid black color pixels to fill the desired size.
Do not modify the original image.
:Parameters:
----------
image : PIL.Image
Input Image sized to fit an input tensor but without padding.
Its possible that one size fits one tensor dimension exactly
but the other size is smaller than
the input tensor other dimension.
desired_size : (width, height)
Exact size expected by the AI model.
:Returns:
-------
PIL.Image
Resized image fitting exactly the AI model input tensor.
"""
assert image
assert desired_size
log.debug("input image size = %r", image.size)
thumb = image.copy()
delta_w = desired_size[0] - thumb.size[0]
delta_h = desired_size[1] - thumb.size[1]
padding = (0, 0, delta_w, delta_h)
new_im = ImageOps.expand(thumb, padding)
log.debug("new image size = %r", new_im.size)
assert new_im.size == desired_size
return new_im
@staticmethod
def resize_to_input_tensor(image=None, desired_size=None):
"""Resize and pad original image to exact size expected by input tensor.
Preserve aspect ratio to avoid confusing the AI model with
distortions that it was not trained on. Pad the resulting image
with solid black color pixels to fill the desired size.
Do not modify anything else in the original image.
:Parameters:
----------
image : PIL.Image
Input Image
desired_size : (width, height)
Exact input tensor size expected by the AI model.
:Returns:
-------
PIL.Image
Resized image fitting exactly the AI model input tensor.
"""
assert image
assert desired_size
# thumbnail is a proportionately resized image
thumbnail = TFDetectionModel.thumbnail(image=image, desired_size=desired_size)
# convert thumbnail into an image with the exact size
# as the input tensor
# preserving proportions by padding as needed
new_im = TFDetectionModel.resize(image=thumbnail, desired_size=desired_size)
return new_im, thumbnail
def log_stats(self, start_time=None):
assert start_time
log.debug("TF engine returned inference results")
end_time = time.monotonic()
inf_time = (end_time - start_time) * 1000
fps = 1.0 / (end_time - self.last_time)
if self.context and self.context.unique_pipeline_name:
pipeline_name = self.context.unique_pipeline_name
else:
pipeline_name = "unknown"
inference_type = type(self).__name__
inf_info = "%s inference time %.2f ms, %.2f fps in pipeline %s"
log.info(inf_info, inference_type, inf_time, fps, pipeline_name)
self.last_time = end_timeClasses
class TFDetectionModel (model=None, labels=None, label_filter=None, confidence_threshold=0.6, top_k=3, **kwargs)-
Applies Tensorflow image detection.
Initialize detector with config parameters. :Parameters:
model: ai_models/mobilenet_ssd_v2_face.tflite labels: ai_models/coco_labels.txt confidence_threshold: 0.6 top_k: 3
Expand source code
class TFDetectionModel(PipeElement): """Applies Tensorflow image detection.""" def __init__( self, model=None, labels=None, label_filter=None, confidence_threshold=0.6, top_k=3, **kwargs, ): """Initialize detector with config parameters. :Parameters: ---------- model: ai_models/mobilenet_ssd_v2_face.tflite labels: ai_models/coco_labels.txt confidence_threshold: 0.6 top_k: 3 """ # log.warning('TFImageDetection __init__ invoked') super().__init__(**kwargs) self._tfengine = TFInferenceEngine( model=model, labels=labels, confidence_threshold=confidence_threshold, top_k=top_k, ) self._labels = self.load_labels(self._tfengine.labels_path) self._label_filter = label_filter self.last_time = time.monotonic() def load_labels(self, label_path=None): """Load label mapping from integer code to text. :Parameters: ---------- label_path : string Path to label mapping file. :Returns: ------- dict {label_code, label_text} """ assert label_path p = re.compile(r"\s*(\d+)(.+)") with open(label_path, encoding="utf-8") as f: lines = (p.match(line).groups() for line in f.readlines()) return {int(num): text.strip() for num, text in lines} @staticmethod def thumbnail(image=None, desired_size=None): """Resizes original image as close as possible to desired size. Preserves aspect ratio of original image. Does not modify the original image. :Parameters: ---------- image : PIL.Image Input Image for AI model detection. desired_size : (width, height) Size expected by the AI model. :Returns: ------- PIL.Image Resized image fitting for the AI model input tensor. """ assert image assert desired_size log.debug("input image size = %r", image.size) thumb = image.copy() w, h = desired_size try: # convert from numpy to native Python int type # that PIL expects if isinstance(w, np.generic): w = w.item() w = int(w) h = h.item() h = int(h) thumb.thumbnail((w, h)) except Exception as e: msg = ( f"Exception in " f"PIL.image.thumbnail(desired_size={desired_size}):" f"type(width)={type(w)}, type(height)={type(h)}" f"\n{e}" ) log.exception(msg) raise RuntimeError(msg) log.debug("thmubnail image size = %r", thumb.size) return thumb @staticmethod def resize(image=None, desired_size=None): """Pad original image to exact size expected by input tensor. Preserve aspect ratio to avoid confusing the AI model with unnatural distortions. Pad the resulting image with solid black color pixels to fill the desired size. Do not modify the original image. :Parameters: ---------- image : PIL.Image Input Image sized to fit an input tensor but without padding. Its possible that one size fits one tensor dimension exactly but the other size is smaller than the input tensor other dimension. desired_size : (width, height) Exact size expected by the AI model. :Returns: ------- PIL.Image Resized image fitting exactly the AI model input tensor. """ assert image assert desired_size log.debug("input image size = %r", image.size) thumb = image.copy() delta_w = desired_size[0] - thumb.size[0] delta_h = desired_size[1] - thumb.size[1] padding = (0, 0, delta_w, delta_h) new_im = ImageOps.expand(thumb, padding) log.debug("new image size = %r", new_im.size) assert new_im.size == desired_size return new_im @staticmethod def resize_to_input_tensor(image=None, desired_size=None): """Resize and pad original image to exact size expected by input tensor. Preserve aspect ratio to avoid confusing the AI model with distortions that it was not trained on. Pad the resulting image with solid black color pixels to fill the desired size. Do not modify anything else in the original image. :Parameters: ---------- image : PIL.Image Input Image desired_size : (width, height) Exact input tensor size expected by the AI model. :Returns: ------- PIL.Image Resized image fitting exactly the AI model input tensor. """ assert image assert desired_size # thumbnail is a proportionately resized image thumbnail = TFDetectionModel.thumbnail(image=image, desired_size=desired_size) # convert thumbnail into an image with the exact size # as the input tensor # preserving proportions by padding as needed new_im = TFDetectionModel.resize(image=thumbnail, desired_size=desired_size) return new_im, thumbnail def log_stats(self, start_time=None): assert start_time log.debug("TF engine returned inference results") end_time = time.monotonic() inf_time = (end_time - start_time) * 1000 fps = 1.0 / (end_time - self.last_time) if self.context and self.context.unique_pipeline_name: pipeline_name = self.context.unique_pipeline_name else: pipeline_name = "unknown" inference_type = type(self).__name__ inf_info = "%s inference time %.2f ms, %.2f fps in pipeline %s" log.info(inf_info, inference_type, inf_time, fps, pipeline_name) self.last_time = end_timeAncestors
Subclasses
Static methods
def resize(image=None, desired_size=None)-
Pad original image to exact size expected by input tensor. Preserve aspect ratio to avoid confusing the AI model with unnatural distortions. Pad the resulting image with solid black color pixels to fill the desired size. Do not modify the original image. :Parameters:
image : PIL.Image Input Image sized to fit an input tensor but without padding. Its possible that one size fits one tensor dimension exactly but the other size is smaller than the input tensor other dimension. desired_size : (width, height) Exact size expected by the AI model. :Returns:
PIL.Image Resized image fitting exactly the AI model input tensor.
Expand source code
@staticmethod def resize(image=None, desired_size=None): """Pad original image to exact size expected by input tensor. Preserve aspect ratio to avoid confusing the AI model with unnatural distortions. Pad the resulting image with solid black color pixels to fill the desired size. Do not modify the original image. :Parameters: ---------- image : PIL.Image Input Image sized to fit an input tensor but without padding. Its possible that one size fits one tensor dimension exactly but the other size is smaller than the input tensor other dimension. desired_size : (width, height) Exact size expected by the AI model. :Returns: ------- PIL.Image Resized image fitting exactly the AI model input tensor. """ assert image assert desired_size log.debug("input image size = %r", image.size) thumb = image.copy() delta_w = desired_size[0] - thumb.size[0] delta_h = desired_size[1] - thumb.size[1] padding = (0, 0, delta_w, delta_h) new_im = ImageOps.expand(thumb, padding) log.debug("new image size = %r", new_im.size) assert new_im.size == desired_size return new_im def resize_to_input_tensor(image=None, desired_size=None)-
Resize and pad original image to exact size expected by input tensor. Preserve aspect ratio to avoid confusing the AI model with distortions that it was not trained on. Pad the resulting image with solid black color pixels to fill the desired size. Do not modify anything else in the original image. :Parameters:
image : PIL.Image Input Image desired_size : (width, height) Exact input tensor size expected by the AI model. :Returns:
PIL.Image Resized image fitting exactly the AI model input tensor.
Expand source code
@staticmethod def resize_to_input_tensor(image=None, desired_size=None): """Resize and pad original image to exact size expected by input tensor. Preserve aspect ratio to avoid confusing the AI model with distortions that it was not trained on. Pad the resulting image with solid black color pixels to fill the desired size. Do not modify anything else in the original image. :Parameters: ---------- image : PIL.Image Input Image desired_size : (width, height) Exact input tensor size expected by the AI model. :Returns: ------- PIL.Image Resized image fitting exactly the AI model input tensor. """ assert image assert desired_size # thumbnail is a proportionately resized image thumbnail = TFDetectionModel.thumbnail(image=image, desired_size=desired_size) # convert thumbnail into an image with the exact size # as the input tensor # preserving proportions by padding as needed new_im = TFDetectionModel.resize(image=thumbnail, desired_size=desired_size) return new_im, thumbnail def thumbnail(image=None, desired_size=None)-
Resizes original image as close as possible to desired size. Preserves aspect ratio of original image. Does not modify the original image. :Parameters:
image : PIL.Image Input Image for AI model detection. desired_size : (width, height) Size expected by the AI model. :Returns:
PIL.Image Resized image fitting for the AI model input tensor.
Expand source code
@staticmethod def thumbnail(image=None, desired_size=None): """Resizes original image as close as possible to desired size. Preserves aspect ratio of original image. Does not modify the original image. :Parameters: ---------- image : PIL.Image Input Image for AI model detection. desired_size : (width, height) Size expected by the AI model. :Returns: ------- PIL.Image Resized image fitting for the AI model input tensor. """ assert image assert desired_size log.debug("input image size = %r", image.size) thumb = image.copy() w, h = desired_size try: # convert from numpy to native Python int type # that PIL expects if isinstance(w, np.generic): w = w.item() w = int(w) h = h.item() h = int(h) thumb.thumbnail((w, h)) except Exception as e: msg = ( f"Exception in " f"PIL.image.thumbnail(desired_size={desired_size}):" f"type(width)={type(w)}, type(height)={type(h)}" f"\n{e}" ) log.exception(msg) raise RuntimeError(msg) log.debug("thmubnail image size = %r", thumb.size) return thumb
Methods
def load_labels(self, label_path=None)-
Load label mapping from integer code to text. :Parameters:
label_path : string Path to label mapping file. :Returns:
dict
Expand source code
def load_labels(self, label_path=None): """Load label mapping from integer code to text. :Parameters: ---------- label_path : string Path to label mapping file. :Returns: ------- dict {label_code, label_text} """ assert label_path p = re.compile(r"\s*(\d+)(.+)") with open(label_path, encoding="utf-8") as f: lines = (p.match(line).groups() for line in f.readlines()) return {int(num): text.strip() for num, text in lines} def log_stats(self, start_time=None)-
Expand source code
def log_stats(self, start_time=None): assert start_time log.debug("TF engine returned inference results") end_time = time.monotonic() inf_time = (end_time - start_time) * 1000 fps = 1.0 / (end_time - self.last_time) if self.context and self.context.unique_pipeline_name: pipeline_name = self.context.unique_pipeline_name else: pipeline_name = "unknown" inference_type = type(self).__name__ inf_info = "%s inference time %.2f ms, %.2f fps in pipeline %s" log.info(inf_info, inference_type, inf_time, fps, pipeline_name) self.last_time = end_time
Inherited members