import zmMagik_helpers.utils as utils
import zmMagik_helpers.globals as g
import zmMagik_helpers.log as log
import cv2
import numpy as np
from shapely.geometry import Polygon
import dateparser
from datetime import datetime, timedelta
import zmMagik_helpers.simpleyolo.simpleYolo as yolo
from ctypes import *
import re
class IMAGE(Structure):
_fields_ = [("w", c_int),
("h", c_int),
("c", c_int),
("data", POINTER(c_float))]
class DetectYolo:
def __init__(self,configPath=None, weightPath=None, labelsPath=None, kernel_fill=3):
if g.args['gpu'] and not g.args['use_opencv_dnn_cuda']:
utils.success_print('Using Darknet GPU model for YOLO')
utils.success_print('If you run out of memory, please tweak yolo.cfg')
if not g.args['use_opencv_dnn_cuda']:
self.m = yolo.SimpleYolo(configPath=configPath,
weightPath=weightPath,
darknetLib=g.args['darknet_lib'],
labelsPath=labelsPath,
useGPU=True)
else:
utils.success_print('Using OpenCV model for YOLO')
utils.success_print('If you run out of memory, please tweak yolo.cfg')
self.net = cv2.dnn.readNetFromDarknet(configPath, weightPath)
self.labels = open(labelsPath).read().strip().split("\n")
np.random.seed(42)
self.colors = np.random.randint(
0, 255, size=(len(self.labels), 3), dtype="uint8")
self.kernel_fill = np.ones((kernel_fill,kernel_fill),np.uint8)
if g.args['use_opencv_dnn_cuda'] and g.args['gpu']:
(maj,minor,patch) = cv2.__version__.split('.')
min_ver = int (maj+minor)
if min_ver < 42:
utils.fail_print('Not setting CUDA backend for OpenCV DNN')
utils.dim_print ('You are using OpenCV version {} which does not support CUDA for DNNs. A minimum of 4.2 is required. See https://www.pyimagesearch.com/2020/02/03/how-to-use-opencvs-dnn-module-with-nvidia-gpus-cuda-and-cudnn/ on how to compile and install openCV 4.2'.format(cv2.__version__))
else:
utils.success_print ('Setting CUDA backend for OpenCV. If you did not set your CUDA_ARCH_BIN correctly during OpenCV compilation, you will get errors during detection related to invalid device/make_policy')
self.net.setPreferableBackend(cv2.dnn.DNN_BACKEND_CUDA)
self.net.setPreferableTarget(cv2.dnn.DNN_TARGET_CUDA)
utils.success_print('YOLO initialized')
def detect(self, frame, frame_b, frame_cnt, orig_fps, starttime, set_frames):
relevant = False
(H, W) = frame.shape[:2]
frame_mask = np.zeros((H, W), dtype=np.uint8)
boxes = []
confidences = []
labels = []
boxed_frame = frame.copy()
if not g.args['gpu'] or g.args['use_opencv_dnn_cuda']:
# we use OpenCV's optimized CPU or GPU code
ln = self.net.getLayerNames()
ln = [ln[i[0] - 1] for i in self.net.getUnconnectedOutLayers()]
blob = cv2.dnn.blobFromImage(
frame, 1 / 255.0, (416, 416), swapRB=True, crop=False)
self.net.setInput(blob)
layerOutputs = self.net.forward(ln)
for output in layerOutputs:
for detection in output:
scores = detection[5:]
classID = np.argmax(scores)
confidence = scores[classID]
label = self.labels[classID]
if confidence > g.args['confidence']:
r = re.compile(g.args['detectpattern'])
if not re.match(r, label):
#utils.dim_print('object "{}" does not match "{}"'.format(label, g.args['detectpattern']))
continue
box = detection[0:4] * np.array([W, H, W, H])
(centerX, centerY, width, height) = box.astype("int")
x = int(centerX - (width / 2))
y = int(centerY - (height / 2))
boxes.append([x, y, int(width), int(height)])
confidences.append(float(confidence))
labels.append(label)
idxs = cv2.dnn.NMSBoxes(boxes, confidences, g.args["confidence"], 0.3)
if len(idxs) > 0:
# loop over the indexes we are keeping
for i in idxs.flatten():
# extract the bounding box coordinates
(x, y) = (boxes[i][0], boxes[i][1])
(width, height) = (boxes[i][2], boxes[i][3])
label = labels[i]
confidence = confidences[i]
pts = Polygon([[x,y], [x+width,y], [x+width, y+height], [x,y+height]])
if g.poly_mask is None or g.poly_mask.intersects(pts):
relevant = True
boxes.append([x, y, int(width), int(height)])
confidences.append(float(confidence))
labels.append(label)
color = (255,0,0)
cv2.rectangle(boxed_frame, (x, y), (x + width, y + height), color, 2)
text = "{}: {:.2f}".format(label, confidence)
cv2.putText(boxed_frame, text, (x, y - 5), cv2.FONT_HERSHEY_SIMPLEX,0.5, color, 2)
if g.args['drawboxes']:
cv2.rectangle(frame_b, (x, y), (x + width, y + height), (255,255,255), 1)
obj_info = {
'name': label,
'time':int(frame_cnt/orig_fps),
'frame': frame_cnt,
'location': ((x,y),(x+width, y+height)),
'confidence': '{:.4f}'.format(confidence)
}
# form text
text = '{}: {}s, Frame: {}'.format(label, int(frame_cnt/orig_fps), frame_cnt)
if starttime:
st = dateparser.parse(starttime)
#from_time = to_time - datetime.timedelta(hours = 1)
# print (st)
dt = st + timedelta(seconds=int(frame_cnt/orig_fps))
text = label + ':' +dt.strftime('%b %d, %I:%M%p')
obj_info['time'] = text
set_frames['frames'].append (obj_info)
# work on displaying text properly
text = text.upper()
delta = 0
d_x = max (x-delta, 0)
d_y = max (y-delta, 0)
d_w = min (W, width+delta)
d_h = min (H, height+delta)
bsx, bsy, bex, bey = utils.write_text(frame=frame_b, text=text, x=d_x, y=d_y, W=W, H=H, adjust=True)
# frame mask of text
#cv2.rectangle(frame_mask, (bsx, bsy), (bex, bey), (255, 255, 255), cv2.FILLED)
# frame mask of object
cv2.rectangle(frame_mask, (d_x,d_y), (d_x+d_w, d_y+d_h), (255, 255, 255), cv2.FILLED)
else: # darknet GPU code
# we use darknet directly
# if you haven't conmpiled darknet in gpu mode, you are going
# to see terrible performance
im = self.m.array_to_image(frame)
detections = self.m.detect_image(im)
boxes = []
confidences = []
labels =[]
for detect in detections:
(label, confidence, bbox) = detect
if confidence > g.args['confidence']:
r = re.compile(g.args['detectpattern'])
if not re.match(r, label):
# utils.dim_print('object "{}" does not match "{}"'.format(label, g.args['detectpattern']))
continue
box = bbox
(centerX, centerY, width, height) = box
x = int(centerX - (width / 2))
y = int(centerY - (height / 2))
width = int(width)
height = int(height)
pts = Polygon([[x,y], [x+width,y], [x+width, y+height], [x,y+height]])
if g.poly_mask is None or g.poly_mask.intersects(pts):
relevant = True
boxes.append([x, y, width, height])
confidences.append(float(confidence))
labels.append(label)
color = (255,0,0)
cv2.rectangle(boxed_frame, (x, y), (x + width, y + height), color, 2)
text = "{}: {:.2f}".format(label, confidence)
cv2.putText(boxed_frame, text, (x, y - 5), cv2.FONT_HERSHEY_SIMPLEX,0.5, color, 2)
if g.args['drawboxes']:
cv2.rectangle(frame_b, (x, y), (x + width, y + height), (255,255,255), 1)
obj_info = {
'name': label,
'time':int(frame_cnt/orig_fps),
'frame': frame_cnt,
'location': ((x,y),(x+width, y+height)),
'confidence': '{:.4f}'.format(confidence)
}
# form text
text = '{}: {}s, Frame: {}'.format(label, int(frame_cnt/orig_fps), frame_cnt)
if starttime:
st = dateparser.parse(starttime)
#from_time = to_time - datetime.timedelta(hours = 1)
# print (st)
dt = st + timedelta(seconds=int(frame_cnt/orig_fps))
text = label + ':' +dt.strftime('%b %d, %I:%M%p')
obj_info['time'] = text
set_frames['frames'].append (obj_info)
# work on displaying text properly
text = text.upper()
delta = 0
d_x = max (x-delta, 0)
d_y = max (y-delta, 0)
d_w = min (W, width+delta)
d_h = min (H, height+delta)
bsx, bsy, bex, bey = utils.write_text(frame=frame_b, text=text, x=d_x, y=d_y, W=W, H=H, adjust=True)
# frame mask of text
#cv2.rectangle(frame_mask, (bsx, bsy), (bex, bey), (255, 255, 255), cv2.FILLED)
# frame mask of object
cv2.rectangle(frame_mask, (d_x,d_y), (d_x+d_w, d_y+d_h), (255, 255, 255), cv2.FILLED)
foreground_a = cv2.bitwise_and(frame,frame, mask=frame_mask)
foreground_b = cv2.bitwise_and(frame_b,frame_b, mask=frame_mask)
combined_fg= cv2.addWeighted(foreground_b, 0.5, foreground_a, 0.5,0)
frame_mask_inv = cv2.bitwise_not(frame_mask)
# blend frame with foreground a missing
modified_frame_b = cv2.bitwise_and(frame_b, frame_b, mask=frame_mask_inv)
merged_frame = cv2.add(modified_frame_b, combined_fg)
# draw mask on blend frame
cv2.polylines(merged_frame, [g.raw_poly_mask], True, (0,0,255), thickness=1)
#return merged_frame, foreground_a, frame_mask, relevant
return merged_frame, foreground_a, frame_mask, relevant, boxed_frame
