def test_fall_detection_case_2_2():
"""Expected to detect a fall because key-points are detected, the angle criteria is met and the time distance between frames is not too short."""
config = _fall_detect_config()
result = None
def sample_callback(image=None, inference_result=None, **kwargs):
nonlocal result
result = inference_result
fall_detector = FallDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
img_1 = _get_image(file_name='fall_img_1.png')
img_2 = _get_image(file_name='fall_img_2.png')
fall_detector.receive_next_sample(image=img_1)
fall_detector.min_time_between_frames = 0.01
time.sleep(fall_detector.min_time_between_frames)
fall_detector.receive_next_sample(image=img_2)
assert result
assert len(result) == 1
category, confidence, box, angle = result[0]
assert box # Add this line to avoid 'Unused local variable'
assert category == 'FALL'
assert confidence > 0.7
assert angle > 60
def test_bad_sample_good_sample():
"""One bad sample should not prevent good samples from being processed."""
config = _fall_detect_config()
result = 'nothing passed to me'
def sample_callback(image=None, inference_result=None, **kwargs):
nonlocal result
result = inference_result
object_detector = ObjectDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
object_detector.connect_to_next_element(output)
# bad sample
object_detector.receive_next_sample(image=None)
assert result == 'nothing passed to me'
# good sample
fall_detector = FallDetector(**config)
fall_detector.connect_to_next_element(output)
img_1 = _get_image(file_name='fall_img_1.png')
img_2 = _get_image(file_name='fall_img_2.png')
fall_detector.receive_next_sample(image=img_1)
fall_detector.min_time_between_frames = 0.01
time.sleep(fall_detector.min_time_between_frames)
fall_detector.receive_next_sample(image=img_2)
assert result
assert len(result) == 1
category, confidence, box, angle = result[0]
assert box # Add this line to avoid 'Unused local variable'
assert category == 'FALL'
assert confidence > 0.7
assert angle > 60
def test_one_person():
"""Expect to detect a fall."""
config = _fall_detect_config()
result = None
def sample_callback(image=None, inference_result=None, **kwargs):
nonlocal result
result = inference_result
fall_detector = FallDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
img_1 = _get_image(file_name='basic_fall_1.jpg')
img_2 = _get_image(file_name='basic_fall_2.jpg')
fall_detector.receive_next_sample(image=img_1)
fall_detector.receive_next_sample(image=img_2)
assert result
assert len(result) == 1
category, confidence, (x0, y0, x1, y1) = result[0]
assert category == 'FALL'
assert confidence > 0.7
assert x0 > 0 and x0 < x1
assert y0 > 0 and y0 < y1
def test_fall_detection_case_2_1():
"""Expected to not detect a fall even though key-points are detected
and the angle criteria is met. However the time distance between
frames is too short."""
config = _fall_detect_config()
result = None
def sample_callback(image=None, inference_result=None, **kwargs):
nonlocal result
result = inference_result
fall_detector = FallDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
# The frame represents a person who is in a standing position.
img_1 = _get_image(file_name="fall_img_1.png")
# The frame represents a person falls.
img_2 = _get_image(file_name="fall_img_2.png")
start_time = time.monotonic()
fall_detector.receive_next_sample(image=img_1)
end_time = time.monotonic()
safe_min = end_time - start_time + 1
# set min time to a sufficiently big number to ensure test passes
# on slow environments
# the goal is to simulate two frames that are too close in time
# to be considered for a fall detection sequence
fall_detector.min_time_between_frames = safe_min
fall_detector.receive_next_sample(image=img_2)
assert not result
def test_fall_detection_case_5():
"""Expected to not detect a fall even the angle criteria is met
because image 2 is standing up rather than fall"""
config = _fall_detect_config()
result = None
def sample_callback(image=None, inference_result=None, **kwargs):
nonlocal result
result = inference_result
fall_detector = FallDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
# The frame represents a person falls.
img_1 = _get_image(file_name="fall_img_2.png")
# The frame represents a person who is in a standing position.
img_2 = _get_image(file_name="fall_img_1.png")
fall_detector.receive_next_sample(image=img_1)
fall_detector.min_time_between_frames = 0.01
time.sleep(fall_detector.min_time_between_frames)
fall_detector.receive_next_sample(image=img_2)
assert not result
def test_fall_detection_case_8():
"""Expect to not detect a fall"""
config = _fall_detect_config()
result = None
def sample_callback(image=None, inference_result=None, **kwargs):
nonlocal result
result = inference_result
fall_detector = FallDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
# No person in a frame
img_1 = _get_image(file_name="fall_img_6.png")
# The frame represents a person who is in a standing position.
img_2 = _get_image(file_name="fall_img_7.png")
fall_detector.receive_next_sample(image=img_1)
# set min time to a small number to speed up testing
fall_detector.min_time_between_frames = 0.01
time.sleep(fall_detector.min_time_between_frames)
fall_detector.receive_next_sample(image=img_2)
assert not result
def test_fall_detection_case_3():
"""Expect to detect a fall as key-points are detected by rotating the image."""
config = _fall_detect_config()
result = None
def sample_callback(image=None, inference_result=None, **kwargs):
nonlocal result
result = inference_result
fall_detector = FallDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
img_1 = _get_image(file_name='fall_img_11.png')
img_2 = _get_image(file_name='fall_img_12.png')
fall_detector.receive_next_sample(image=img_1)
# set min time to a small number to speed up testing
fall_detector.min_time_between_frames = 0.01
time.sleep(fall_detector.min_time_between_frames)
fall_detector.receive_next_sample(image=img_2)
assert result
assert len(result) == 1
category, confidence, box, angle = result[0]
assert box # Add this line to avoid 'Unused local variable'
assert category == 'FALL'
assert confidence > 0.3
assert angle > 60
def test_fall_detection_case_1():
"""Expected to not detect a fall as key-points are not detected."""
config = _fall_detect_config()
result = None
def sample_callback(image=None, inference_result=None, **kwargs):
nonlocal result
result = inference_result
fall_detector = FallDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
# The frame represents a person who is in a standing position.
img_1 = _get_image(file_name="fall_img_1.png")
# The frame represents a person completely falls.
img_2 = _get_image(file_name="fall_img_3.png")
fall_detector.receive_next_sample(image=img_1)
fall_detector.min_time_between_frames = 0.01
time.sleep(fall_detector.min_time_between_frames)
fall_detector.receive_next_sample(image=img_2)
assert not result
def test_config_confidence_threshold():
"""Verify against known confidence threshold. Make sure it propagates
at all levels."""
config = _fall_detect_config()
fall_detector = FallDetector(**config)
tfe = fall_detector._tfengine
pe = fall_detector._pose_engine
assert fall_detector.confidence_threshold == config["confidence_threshold"]
assert pe.confidence_threshold == config["confidence_threshold"]
assert tfe.confidence_threshold == config["confidence_threshold"]
config["confidence_threshold"] = 0.457
fall_detector = FallDetector(**config)
tfe = fall_detector._tfengine
pe = fall_detector._pose_engine
assert fall_detector.confidence_threshold == config["confidence_threshold"]
assert pe.confidence_threshold == config["confidence_threshold"]
assert tfe.confidence_threshold == config["confidence_threshold"]
def test_draw_line_1_1():
"""One keypoing but no full body line. No image should be saved."""
config = _fall_detect_config()
fall_detector = FallDetector(**config)
image = _get_image(file_name="fall_img_1.png")
pose_dix = {fall_detector.LEFT_SHOULDER: [0, 0]}
lines_drawn = fall_detector.draw_lines(image, pose_dix, 0.5)
assert lines_drawn == 0
def test_fall_detection_2_frame_back_case_1():
"""
Expected to detect a fall using frame[t] and frame[t-1].
frame[t-2] : A person is in standing position.
frame[t-1] : A person is almost in standing position as he is walking.
frame[t] : A person is fall down.
"""
config = _fall_detect_config()
result = None
def sample_callback(image=None, inference_result=None, **kwargs):
nonlocal result
result = inference_result
fall_detector = FallDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
# A frame at t-2 timestamp when person is in standing position.
img_1 = _get_image(file_name="fall_img_1.png")
# A frame at t-1 timestamp when person is almost in standing position \
# as he is walking.
img_2 = _get_image(file_name="fall_img_1_1.png")
# A frame at t timestamp when person falls down.
img_3 = _get_image(file_name="fall_img_2.png")
fall_detector.min_time_between_frames = 0.01
fall_detector.receive_next_sample(image=img_1)
time.sleep(fall_detector.min_time_between_frames)
fall_detector.receive_next_sample(image=img_2)
time.sleep(fall_detector.min_time_between_frames)
assert not result
fall_detector.receive_next_sample(image=img_3)
assert result
assert len(result) == 1
category = result[0]["label"]
confidence = result[0]["confidence"]
angle = result[0]["leaning_angle"]
keypoint_corr = result[0]["keypoint_corr"]
assert keypoint_corr
assert category == "FALL"
assert confidence > 0.7
assert angle > 60
def test_fall_detection_2_frame_back_case_2():
"""
Expected to detect a fall using frame[t] and frame[t-2].
frame[t-2] : A person is in standing position.
frame[t-1] : A person is mid-way of fall.
frame[t] : A person is fall down.
"""
config = _fall_detect_config()
result = None
def sample_callback(image=None, inference_result=None, **kwargs):
nonlocal result
result = inference_result
fall_detector = FallDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
# A frame at t-2 timestamp when person is in standing position.
img_1 = _get_image(file_name='fall_img_1.png')
# A frame at t-1 timestamp when person is mid-way of fall.
img_2 = _get_image(file_name='fall_img_2_2.png')
# A frame at t timestamp when person falls down.
img_3 = _get_image(file_name='fall_img_2.png')
fall_detector.min_time_between_frames = 0.01
fall_detector.max_time_between_frames = 15
fall_detector.receive_next_sample(image=img_1)
time.sleep(fall_detector.min_time_between_frames)
fall_detector.receive_next_sample(image=img_2)
time.sleep(fall_detector.min_time_between_frames)
assert not result
fall_detector.receive_next_sample(image=img_3)
assert result
assert len(result) == 1
category = result[0]['label']
confidence = result[0]['confidence']
angle = result[0]['leaning_angle']
keypoint_corr = result[0]['keypoint_corr']
assert keypoint_corr
assert category == 'FALL'
assert confidence > 0.7
assert angle > 60
def test_draw_line_1():
"""One body line passed to draw. Image with one line should be saved."""
config = _fall_detect_config()
fall_detector = FallDetector(**config)
image = _get_image(file_name="fall_img_1.png")
pose_dix = {
fall_detector.LEFT_SHOULDER: [0, 0],
fall_detector.LEFT_HIP: [0, 1]
}
lines_drawn = fall_detector.draw_lines(image, pose_dix, 0.5)
assert lines_drawn == 1
def test_model_inputs():
"""Verify against known model inputs."""
config = _fall_detect_config()
fall_detector = FallDetector(**config)
tfe = fall_detector._tfengine
samples = tfe.input_details[0]['shape'][0]
assert samples == 1
height = tfe.input_details[0]['shape'][1]
assert height == 300
width = tfe.input_details[0]['shape'][2]
assert width == 300
colors = tfe.input_details[0]['shape'][3]
assert colors == 3
def test_draw_line_0():
"""No body lines passed to draw. No image should be saved."""
config = _fall_detect_config()
fall_detector = FallDetector(**config)
image = _get_image(file_name="fall_img_1.png")
pose_dix = None
lines_drawn = fall_detector.draw_lines(image, pose_dix, 0.5)
assert lines_drawn == 0
pose_dix = {}
lines_drawn = fall_detector.draw_lines(image, pose_dix, 0.5)
assert lines_drawn == 0
def test_no_sample():
"""Expect element to pass empty sample to next element."""
config = _fall_detect_config()
result = False
def sample_callback(image=None, inference_result=None, **kwargs):
nonlocal result
result = image is None and inference_result is None
fall_detector = FallDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
fall_detector.receive_next_sample()
assert result is True
def test_background_image():
"""Expect to not detect anything interesting in a background image."""
config = _fall_detect_config()
result = None
def sample_callback(image=None, inference_result=None, **kwargs):
nonlocal result
result = inference_result
fall_detector = FallDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
img = _get_image(file_name='background.jpg')
fall_detector.receive_next_sample(image=img)
assert not result
def test_model_outputs():
"""Verify against known model outputs."""
config = _fall_detect_config()
object_detector = FallDetector(**config)
tfe = fall_detector._tfengine
assert tfe.output_details[0]['shape'][0] == 1
scores = tfe.output_details[0]['shape'][1]
assert scores == 20
assert tfe.output_details[1]['shape'][0] == 1
boxes = tfe.output_details[1]['shape'][1]
assert boxes == 20
assert tfe.output_details[2]['shape'][0] == 1
labels = tfe.output_details[2]['shape'][1]
assert labels == 20
num = tfe.output_details[3]['shape'][0]
assert num == 1
def test_draw_line_2():
"""Two body lines passed to draw. Image with two lines should be saved."""
config = _fall_detect_config()
fall_detector = FallDetector(**config)
# The frame represents a person who is in a standing position.
image = _get_image(file_name="fall_img_1.png")
pose_dix = {
fall_detector.LEFT_SHOULDER: [0, 0],
fall_detector.LEFT_HIP: [0, 1],
fall_detector.RIGHT_SHOULDER: [1, 0],
fall_detector.RIGHT_HIP: [1, 1],
}
lines_drawn = fall_detector.draw_lines(image, pose_dix, 0.5)
assert lines_drawn == 2
def test_bad_sample_good_sample():
"""One bad sample should not prevent good samples from being processed."""
config = _fall_detect_config()
result = "nothing passed to me"
def sample_callback(image=None, inference_result=None, **kwargs):
nonlocal result
result = inference_result
object_detector = ObjectDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
object_detector.connect_to_next_element(output)
# bad sample
object_detector.receive_next_sample(image=None)
assert result == "nothing passed to me"
# good sample
fall_detector = FallDetector(**config)
fall_detector.connect_to_next_element(output)
# The frame represents a person who is in a standing position.
img_1 = _get_image(file_name="fall_img_1.png")
# The frame represents a person falls.
img_2 = _get_image(file_name="fall_img_2.png")
fall_detector.receive_next_sample(image=img_1)
fall_detector.min_time_between_frames = 0.01
time.sleep(fall_detector.min_time_between_frames)
fall_detector.receive_next_sample(image=img_2)
assert result
assert len(result) == 1
category = result[0]["label"]
confidence = result[0]["confidence"]
angle = result[0]["leaning_angle"]
keypoint_corr = result[0]["keypoint_corr"]
assert keypoint_corr
assert category == "FALL"
assert confidence > 0.7
assert angle > 60
def test_fall_detection_2_frame_back_case_3():
"""
Expected to not detect a fall using frame[t],frame[t-1] and frame[t-2].
frame[t-2] : A person is in walking postion.
frame[t-1] : A person is in walking postion.
frame[t] : A person is slight in lean postion but no fall.
"""
config = _fall_detect_config()
result = None
def sample_callback(image=None, inference_result=None, **kwargs):
nonlocal result
result = inference_result
fall_detector = FallDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
# A frame at t-2 timestamp when person is in walking postion.
img_1 = _get_image(file_name="fall_img_15.png")
# A frame at t-1 timestamp when person is in walking postion.
img_2 = _get_image(file_name="fall_img_16.png")
# A frame at t timestamp when person is slight in lean postion but no fall.
img_3 = _get_image(file_name="fall_img_17.png")
fall_detector.min_time_between_frames = 0.01
fall_detector.receive_next_sample(image=img_1)
time.sleep(fall_detector.min_time_between_frames)
fall_detector.receive_next_sample(image=img_2)
time.sleep(fall_detector.min_time_between_frames)
assert not result
fall_detector.receive_next_sample(image=img_3)
assert not result
def test_fall_detection_thumbnail_present():
"""Expected to receive thumnail in result if image is provided \
and poses are detected."""
config = _fall_detect_config()
result = None
def sample_callback(image=None,
thumbnail=None,
inference_result=None,
**kwargs):
nonlocal result
result = (image is not None and thumbnail is not None
and inference_result is not None)
fall_detector = FallDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
img_1 = _get_image(file_name="fall_img_1.png")
fall_detector.receive_next_sample(image=img_1)
assert result is True
def test_fall_detection_case_3_2():
"""Expect to detect a fall as key-points are detected
by rotating the image counter clockwise."""
config = _fall_detect_config()
result = None
def sample_callback(image=None, inference_result=None, **kwargs):
nonlocal result
result = inference_result
fall_detector = FallDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
# The frame represents a person who is in a standing position.
img_1 = _get_image(file_name="fall_img_11_flip.png")
# The frame represents a person completely falls.
img_2 = _get_image(file_name="fall_img_12_flip.png")
fall_detector.receive_next_sample(image=img_1)
# set min time to a small number to speed up testing
fall_detector.min_time_between_frames = 0.01
time.sleep(fall_detector.min_time_between_frames)
fall_detector.receive_next_sample(image=img_2)
assert result
assert len(result) == 1
category = result[0]["label"]
confidence = result[0]["confidence"]
angle = result[0]["leaning_angle"]
keypoint_corr = result[0]["keypoint_corr"]
assert keypoint_corr
assert category == "FALL"
assert confidence > 0.3
assert angle > 60
def test_fall_detection_case_2_2():
"""Expected to detect a fall because key-points are detected,
the angle criteria is met and the time distance between
frames is not too short."""
config = _fall_detect_config()
result = None
def sample_callback(image=None, inference_result=None, **kwargs):
nonlocal result
result = inference_result
fall_detector = FallDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
# The frame represents a person who is in a standing position.
img_1 = _get_image(file_name="fall_img_1.png")
# The frame represents a person falls.
img_2 = _get_image(file_name="fall_img_2.png")
fall_detector.receive_next_sample(image=img_1)
fall_detector.min_time_between_frames = 0.01
time.sleep(fall_detector.min_time_between_frames)
fall_detector.receive_next_sample(image=img_2)
assert result
assert len(result) == 1
category = result[0]["label"]
confidence = result[0]["confidence"]
angle = result[0]["leaning_angle"]
keypoint_corr = result[0]["keypoint_corr"]
assert keypoint_corr
assert category == "FALL"
assert confidence > 0.7
assert angle > 60
def test_background_image():
"""Expect to not detect anything interesting in a background image."""
config = _fall_detect_config()
result = None
def sample_callback(image=None,
thumbnail=None,
inference_result=None,
**kwargs):
nonlocal result
result = image is not None and thumbnail is not None and not inference_result
fall_detector = FallDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
img = _get_image(file_name="background.jpg")
fall_detector.receive_next_sample(image=img)
fall_detector.min_time_between_frames = 0.01
time.sleep(fall_detector.min_time_between_frames)
img = _get_image(file_name="background.jpg")
fall_detector.receive_next_sample(image=img)
assert result is True
def test_fall_detection_case_4():
"""No Fall"""
config = _fall_detect_config()
result = None
def sample_callback(image=None, inference_result=None, **kwargs):
nonlocal result
result = inference_result
fall_detector = FallDetector(**config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
img_1 = _get_image(file_name='fall_img_1.png')
img_2 = _get_image(file_name='fall_img_4.png')
fall_detector.receive_next_sample(image=img_1)
fall_detector.min_time_between_frames = 0.01
time.sleep(fall_detector.min_time_between_frames)
fall_detector.receive_next_sample(image=img_2)
assert not result
def _helper_test_debug_image_save(context: PipelineContext = None):
log_config = {"level": "DEBUG"}
logger.configure(config=log_config)
# Expect to receive thumnail in result if image is provided and
# poses are detected.
config = _fall_detect_config()
result = None
def sample_callback(image=None,
thumbnail=None,
inference_result=None,
**kwargs):
nonlocal result
result = (image is not None and thumbnail is not None
and inference_result is not None)
fall_detector = FallDetector(context=context, **config)
output = _OutPipeElement(sample_callback=sample_callback)
fall_detector.connect_to_next_element(output)
img_1 = _get_image(file_name="fall_img_1.png")
fall_detector.receive_next_sample(image=img_1)
assert result is True
# now that we know there was a positive detection
# lets check if the the interim debug images were saved as expected
pose_img_files = list(_data_dir.glob("tmp-pose-detect-image*.jpg"))
assert len(pose_img_files) > 0
fall_img_files = list(_data_dir.glob("tmp-fall-detect-thumbnail*.jpg"))
assert len(fall_img_files) > 0
# cleanup after test
all_tmp_files = pose_img_files + fall_img_files
for f in all_tmp_files:
f.unlink()
# return logger level to INFO to prevent side effects in other tests
log_config = {"level": "INFO"}
logger.configure(config=log_config)