Skip to content

Add TensorFlow Edge Impulse #118

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
IAmCoder merged 2 commits into from
Mar 31, 2025
Merged

Add TensorFlow Edge Impulse #118

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
79b6095
Add angle correction
IAmCoder Mar 31, 2025
f724d89
Add Edge Impulse Faster Objects More Objects
IAmCoder Mar 31, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Binary file modified release/firmware.bin
View file
Open in desktop
Binary file not shown.
1 change: 1 addition & 0 deletions software/config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,6 +35,7 @@ def __init__(self):
self.default['CreateGifs'] = 1
self.default['TrackFace'] = 1
self.default['TrackEyes'] = 0
self.default['TensorFlow'] = 1
self.default['FaceFeatures'] = 16
self.default['DrawFaceRegion'] = 1
self.default['FaceStages'] = 25
Expand Down
100 changes: 90 additions & 10 deletions software/face.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,8 @@
import sensor
import image
import utime
import math
import ml
from ml.utils import NMS

class face_detection:
def __init__(self, config, comms):
Expand All @@ -9,34 +11,63 @@ def __init__(self, config, comms):
self.has_face = False
self.face_cascade = image.HaarCascade("frontalface", stages=self.config.get('FaceStages'))
self.face_object = [0, 0, 1, 1]
self.face_angle = 0
self.correct_angle = False
self.ml_model = None
self.min_confidence = 0.4
self.threshold_list = [(math.ceil(self.min_confidence * 255), 255)]

self.extra_fb = sensor.alloc_extra_fb(sensor.width(), sensor.height(), sensor.GRAYSCALE)

def detect(self, img):
def detect(self, img, global_variance):
if not self.config.get('TrackFace'):
self.face_object = [0, 0, img.width(), img.height()]
self.has_face = False
return

now = utime.ticks_ms()
if global_variance >= self.config.get('TossThreshold'):
self.face_object = [0, 0, img.width(), img.height()]
self.has_face = False

if self.config.get('TensorFlow'):
if self.ml_model == None:
self.ml_model = ml.Model('fomo_face_detection')

for i, detection_list in enumerate(self.ml_model.predict([img], callback=self.post_process)):
if i == 0 or len(detection_list) == 0:
continue

for (x, y, w, h), score in detection_list:
self.face_object = (x - w, y - h, w * 4, h * 4)
self.has_face = True
return

self.face_angle = 0
face_objects = img.find_features(self.face_cascade, threshold=self.config.get('FaceThreshold'), scale_factor=self.config.get('FaceScaleFactor'))

if len(face_objects) == 0 and self.config.get('FaceAngles'):
for angle in self.config.get('FaceAngles'):
self.extra_fb.replace(img)
self.extra_fb.rotation_corr(x_rotation=0.0, y_rotation=0.0, z_rotation=angle)

face_objects = self.extra_fb.find_features(self.face_cascade, threshold=self.config.get('FaceThreshold'), scale_factor=self.config.get('FaceScaleFactor'))
if face_objects:
self.face_angle = angle
break

if face_objects:
self.face_object = face_objects[0]
self.has_face = True

if self.config.get('TrackEyes'):
self.face_object = [self.face_object[0], self.face_object[1] + int(self.face_object[3] * 1/5), self.face_object[2], int(self.face_object[3] * 2/5)]
else:
self.face_object = [0, 0, img.width(), img.height()]
self.has_face = False

if self.correct_angle:
eyes_x = self.face_object[0] + int(self.face_object[2] * 1/5)
eyes_y = self.face_object[1] + int(self.face_object[3] * 1/5)
eyes_width = self.face_object[2] - int(self.face_object[2] * 1/5)
eyes_height = int(self.face_object[3] * 2/5)
self.face_object = [eyes_x, eyes_y, eyes_width, eyes_height]

def draw_region(self, img):
if not self.config.get('DrawFaceRegion'):
Expand All @@ -45,7 +76,56 @@ def draw_region(self, img):
if self.face_object[2] == img.width():
return

if self.config.get('PixelFormat') == 'RGB565':
img.draw_rectangle(self.face_object, color=(70, 130, 180))
else:
img.draw_rectangle(self.face_object)
img.draw_rectangle(self.face_object, color=(70, 130, 180))

if self.correct_angle:
face_x = self.face_object[0]
face_y = self.face_object[1]
face_width = self.face_object[2]
face_height = self.face_object[3]

theta = math.radians(self.face_angle * -1)
offset = (0, 0)

rect = [
(face_x, face_y),
(face_x, face_y + face_height),
(face_x + face_width, face_y + face_height),
(face_x + face_width, face_y)
]

rotated_rect = [translate(rotate(xy, theta), offset) for xy in rect]

img.draw_line((rotated_rect[0][0], rotated_rect[0][1], rotated_rect[1][0], rotated_rect[1][1]), color=(220, 220, 0))
img.draw_line((rotated_rect[1][0], rotated_rect[1][1], rotated_rect[2][0], rotated_rect[2][1]), color=(220, 220, 0))
img.draw_line((rotated_rect[2][0], rotated_rect[2][1], rotated_rect[3][0], rotated_rect[3][1]), color=(220, 220, 0))
img.draw_line((rotated_rect[3][0], rotated_rect[3][1], rotated_rect[0][0], rotated_rect[0][1]), color=(220, 220, 0))
img.draw_string(face_x, face_y + face_height - 10, str(self.face_angle) + "^", color=(70, 130, 180), mono_space=False)

def post_process(self, model, inputs, outputs):
n, oh, ow, oc = model.output_shape[0]
nms = NMS(ow, oh, inputs[0].roi)
for i in range(oc):
img = image.Image(outputs[0][0, :, :, i] * 255)
blobs = img.find_blobs(
self.threshold_list, x_stride=1, area_threshold=1, pixels_threshold=1
)
for b in blobs:
rect = b.rect()
x, y, w, h = rect
score = (
img.get_statistics(thresholds=self.threshold_list, roi=rect).l_mean() / 255.0
)
nms.add_bounding_box(x, y, x + w, y + h, score, i)
return nms.get_bounding_boxes()

def rotate(xy, theta):
cos_theta, sin_theta = math.cos(theta), math.sin(theta)

return (
int(xy[0] * cos_theta - xy[1] * sin_theta),
int(xy[0] * sin_theta + xy[1] * cos_theta)
)

def translate(xy, offset):
return xy[0] + offset[0], xy[1] + offset[1]
7 changes: 1 addition & 6 deletions software/inspec.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -98,12 +98,7 @@ def monitor(self):
try:
self.img = sensor.snapshot()

if self.config.get('TrackFace'):
self.img.gamma(gamma=1.0, contrast=1.5, brightness=0.0)
else:
time.sleep_ms(64)

self.face.detect(self.img)
self.face.detect(self.img, self.global_variance)
self.global_variance, self.variance = self.img.variance(self.extra_fb, self.config.get('PixelThreshold'), self.config.get('PixelRange'), self.face.face_object)

if self.variance > self.peak_variance:
Expand Down
1 change: 1 addition & 0 deletions software/rem.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,7 @@ def __init__(self, config, face):
def detect(self, variance, global_variance):
if global_variance >= self.config.get('TossThreshold'):
self.eye_movements = 0
self.last_eye_movement = now + 1000 * 10

now = utime.ticks_ms()
if now - self.last_eye_movement > 1000 * 60:
Expand Down
2 changes: 1 addition & 1 deletion software/version.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1 +1 @@
version = '1.2.5'
version = '1.2.6'