def getBoundingBoxes(self, image, threshold=0.8):
bodies = self.model.detect([self.preprocessImage(image)], verbose=0)[0]
boxes = []
for id, score, box in zip(bodies["class_ids"], bodies["scores"],
bodies["rois"]):
if id == 1 and score > threshold:
boxes.append(
BoundingBox(Vector(int(box[1]), int(box[0])),
Vector(int(box[3]), int(box[2]))))
return image, boxes
def getBoundingBoxesAndMasks(self, image, threshold=0.8):
bodies = self.model.detect([self.preprocessImage(image)], verbose=0)[0]
boxes = []
masks = []
for index, (id, score, box) in enumerate(
zip(bodies["class_ids"], bodies["scores"], bodies["rois"])):
if id == 1 and score > threshold:
boxes.append(
BoundingBox(Vector(int(box[1]), int(box[0])),
Vector(int(box[3]), int(box[2]))))
masks.append(bodies["masks"][:, :, index])
return (boxes, masks)
def getBoundingBoxes(self, image, threshold=0.5):
blob = cv.dnn.blobFromImage(image,
1 / 255.,
self.inputDim,
swapRB=False,
crop=False)
self.model.setInput(blob)
outs = self.model.forward(self.getOutputLayers())
boxVectors = []
confidence = []
for out in outs:
for detection in out:
id = np.argmax(detection[5:])
score = detection[5:][id]
if id == 0 and score > threshold:
boxVector = detection[:4] * np.array([
image.shape[1], image.shape[0], image.shape[1],
image.shape[0]
])
boxVectors.append([
int(boxVector[0] - (boxVector[2] / 2)),
int(boxVector[1] - (boxVector[3] / 2)),
int(boxVector[2]),
int(boxVector[3])
])
confidence.append(float(score))
boxes = []
if boxVectors:
indexes = cv.dnn.NMSBoxes(boxVectors, confidence, threshold,
self.nmsThreshold)
boxes = [
BoundingBox(
Vector(boxVector[0], boxVector[1]),
Vector(boxVector[0] + boxVector[2],
boxVector[1] + boxVector[3]))
for boxVector in map(lambda i: boxVectors[i[0]], indexes)
]
return image, boxes
def add_vector(self, vid, vector_array):
self.vectors.append(Vector(vid, vector_array))