def __init__(self, IP, cameraId=vd.kTopCamera, resolution=vd.kVGA):
super(ObjectDetection, self).__init__(IP, cameraId, resolution)
self._boundRect = []
self._cropKeep = 1
self._stickAngle = None # rad
#self._classes_name = ["stick"]
self._common_params = {'image_size': 448, 'num_classes': 1,
'batch_size':1}
self._net_params = {'cell_size': 7, 'boxes_per_cell':2, 'weight_decay': 0.0005}
self._net = YoloTinyNet(self._common_params, self._net_params, test=True)
#self._modelFile = "/home/meringue/Documents/python-nao-golf/yoloNet/models/train/model.ckpt-95000"
#self._objectRect = [0, 0, 0, 0]
self._objectName = None
def __init__(self,
IP,
classes_name,
cameraId=vd.kTopCamera,
resolution=vd.kVGA):
super(MultiObjectDetection, self).__init__(IP, cameraId, resolution)
self._classes_name = classes_name
self._num_classes = len(classes_name)
self._common_params = {
'image_size': 448,
'num_classes': self._num_classes,
'batch_size': 1
}
self._net_params = {
'cell_size': 7,
'boxes_per_cell': 2,
'weight_decay': 0.0005
}
self._net = YoloTinyNet(self._common_params,
self._net_params,
test=True)
def main():
net = YoloTinyNet(common_params, net_params, test=True)
# tensorflow中声明占位符号image, 这在后面run的时候
# feed_dict中会出现该占位符和对应的值，意思就是输入数据的来源
image = tf.placeholder(tf.float32, (1, 448, 448, 3))
predicts = net.inference(image)
sess = tf.Session()
# 转化数据格式
np_img = cv2.imread('cat.jpg')
resized_img = cv2.resize(np_img, (448, 448))
np_img = cv2.cvtColor(resized_img, cv2.COLOR_BGR2RGB)
np_img = np_img.astype(np.float32)
#白化输入的数据
np_img = np_img / 255.0 * 2 - 1
np_img = np.reshape(np_img, (1, 448, 448, 3))
saver = tf.train.Saver(net.trainable_collection)
saver.restore(sess, 'models/pretrain/yolo_tiny.ckpt')
# The optional feed_dict argument allows the caller to override
# the value of tensors in the graph.
np_predict = sess.run(predicts, feed_dict={image: np_img})
xmin, ymin, xmax, ymax, class_num = process_predicts(np_predict)
class_name = classes_name[class_num]
# 绘制预测框, 输出预测类型
cv2.rectangle(resized_img, (int(xmin), int(ymin)), (int(xmax), int(ymax)),
(0, 0, 255))
cv2.putText(resized_img, class_name, (int(xmin), int(ymin)), 2, 1.5,
(0, 0, 255))
cv2.imwrite('cat_out.jpg', resized_img)
sess.close()
src_xmax = xmax * width_ratio
src_ymax = ymax * height_ratio
score = float("%.3f" %score)
cv2.rectangle(src_img, (int(src_xmin), int(src_ymin)), (int(src_xmax), int(src_ymax)), (0, 0, 255))
cv2.putText(src_img, object_name + str(score), (int(src_xmin), int(src_ymin)), 1, 2, (0, 0, 255))
#cv2.imshow("result", src_img)
cv2.imwrite("result.jpg", src_img)
if __name__ == '__main__':
common_params = {'image_size': 448, 'num_classes': 20, 'batch_size': 1}
net_params = {'cell_size': 7, 'boxes_per_cell': 2, 'weight_decay': 0.0005}
net = YoloTinyNet(common_params, net_params, test=True)
image = tf.placeholder(tf.float32, (1, 448, 448, 3))
predicts = net.yoloTinyModel(image)
os.environ["CUDA_VISIBLE_DEVICES"] = "1"
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
sess = tf.Session(config=config)
src_img = cv2.imread("./test2.jpg")
#src_img = cv2.imread("./data/VOCdevkit2007/VOC2007/JPEGImages/000058.jpg")
resized_img = cv2.resize(src_img, (448, 448))
#height_ratio = src_img.shape[0]/448.0
#width_ratio = src_img.shape[1]/448.0
w = w * x_size
h = h * y_size
xmin = xcenter - w / 2.0
ymin = ycenter - h / 2.0
xmax = xmin + w
ymax = ymin + h
return xmin, ymin, xmax, ymax, class_num
common_params = {'image_size': x_size, 'num_classes': 20, 'batch_size': 1}
net_params = {'cell_size': 7, 'boxes_per_cell': 2, 'weight_decay': 0.0005}
net = YoloTinyNet(common_params, net_params, test=True)
image = tf.placeholder(tf.float32, (1, x_size, y_size, channel))
predicts = net.inference(image)
saver = tf.train.Saver(net.trainable_collection)
with tf.Session() as sess:
saver.restore(sess, model_path)
forderlist = os.listdir(directory)
filecnt = 0
for forder in forderlist:
filelist = os.listdir(directory + '/' + forder)
for filename in filelist:
# PNG -> JPEG
img = Image.open(directory + '/' + forder + '/' + filename)
w = w * 448
h = h * 448
xmin = xcenter - w/2.0
ymin = ycenter - h/2.0
xmax = xmin + w
ymax = ymin + h
return xmin, ymin, xmax, ymax, class_num
common_params = {'image_size': 448, 'num_classes': 20,
'batch_size':1}
net_params = {'cell_size': 7, 'boxes_per_cell':2, 'weight_decay': 0.0005}
net = YoloTinyNet(common_params, net_params, test=True)
image = tf.placeholder(tf.float32, (1, 448, 448, 3))
predicts = net.inference(image)
sess = tf.Session()
np_img = cv2.imread('dining_table.jpg')
resized_img = cv2.resize(np_img, (448, 448))
np_img = cv2.cvtColor(resized_img, cv2.COLOR_BGR2RGB)
np_img = np_img.astype(np.float32)
np_img = np_img / 255.0 * 2 - 1
np_img = np.reshape(np_img, (1, 448, 448, 3))
class MultiObjectDetection(VisualBasis):
def __init__(self,
IP,
classes_name,
cameraId=vd.kTopCamera,
resolution=vd.kVGA):
super(MultiObjectDetection, self).__init__(IP, cameraId, resolution)
self._classes_name = classes_name
self._num_classes = len(classes_name)
self._common_params = {
'image_size': 448,
'num_classes': self._num_classes,
'batch_size': 1
}
self._net_params = {
'cell_size': 7,
'boxes_per_cell': 2,
'weight_decay': 0.0005
}
self._net = YoloTinyNet(self._common_params,
self._net_params,
test=True)
def predict_object(self, image):
predicts = self._net.inference(image)
return predicts
def process_predicts(self, resized_img, predicts, thresh=0.2):
"""
process the predicts of object detection with one image input.
Args:
resized_img: resized source image.
predicts: output of the model.
thresh: thresh of bounding box confidence.
Return:
predicts_dict: {"stick": [[x1, y1, x2, y2, scores1], [...]]}.
"""
cls_num = self._num_classes
bbx_per_cell = self._net_params["boxes_per_cell"]
cell_size = self._net_params["cell_size"]
img_size = self._common_params["image_size"]
p_classes = predicts[0, :, :, 0:cls_num]
C = predicts[0, :, :, cls_num:cls_num +
bbx_per_cell] # two bounding boxes in one cell.
coordinate = predicts[0, :, :, cls_num +
bbx_per_cell:] # all bounding boxes position.
p_classes = np.reshape(p_classes, (cell_size, cell_size, 1, cls_num))
C = np.reshape(C, (cell_size, cell_size, bbx_per_cell, 1))
P = C * p_classes # confidencefor all classes of all bounding boxes (cell_size, cell_size, bounding_box_num, class_num) = (7, 7, 2, 1).
predicts_dict = {}
for i in range(cell_size):
for j in range(cell_size):
temp_data = np.zeros_like(P, np.float32)
temp_data[i, j, :, :] = P[i, j, :, :]
position = np.argmax(
temp_data
) # refer to the class num (with maximum confidence) for every bounding box.
index = np.unravel_index(position, P.shape)
if P[index] > thresh:
class_num = index[-1]
coordinate = np.reshape(
coordinate, (cell_size, cell_size, bbx_per_cell, 4)
) # (cell_size, cell_size, bbox_num_per_cell, coordinate)[xmin, ymin, xmax, ymax]
max_coordinate = coordinate[index[0], index[1],
index[2], :]
xcenter = max_coordinate[0]
ycenter = max_coordinate[1]
w = max_coordinate[2]
h = max_coordinate[3]
xcenter = (index[1] + xcenter) * (1.0 * img_size /
cell_size)
ycenter = (index[0] + ycenter) * (1.0 * img_size /
cell_size)
w = w * img_size
h = h * img_size
xmin = 0 if (xcenter - w / 2.0 < 0) else (xcenter -
w / 2.0)
ymin = 0 if (xcenter - w / 2.0 < 0) else (ycenter -
h / 2.0)
xmax = resized_img.shape[0] if (
xmin + w) > resized_img.shape[0] else (xmin + w)
ymax = resized_img.shape[1] if (
ymin + h) > resized_img.shape[1] else (ymin + h)
class_name = self._classes_name[class_num]
predicts_dict.setdefault(class_name, [])
predicts_dict[class_name].append(
[int(xmin),
int(ymin),
int(xmax),
int(ymax), P[index]])
return predicts_dict
def non_max_suppress(self, predicts_dict, threshold=0.5):
"""
implement non-maximum supression on predict bounding boxes.
Args:
predicts_dict: {"stick": [[x1, y1, x2, y2, scores1], [...]]}.
threshhold: iou threshold
Return:
predicts_dict processed by non-maximum suppression
"""
for object_name, bbox in predicts_dict.items():
bbox_array = np.array(bbox, dtype=np.float)
x1, y1, x2, y2, scores = bbox_array[:,
0], bbox_array[:,
1], bbox_array[:,
2], bbox_array[:,
3], bbox_array[:,
4]
areas = (x2 - x1 + 1) * (y2 - y1 + 1)
order = scores.argsort()[::-1]
keep = []
while order.size > 0:
i = order[0]
keep.append(i)
xx1 = np.maximum(x1[i], x1[order[1:]])
yy1 = np.maximum(y1[i], y1[order[1:]])
xx2 = np.minimum(x2[i], x2[order[1:]])
yy2 = np.minimum(y2[i], y2[order[1:]])
inter = np.maximum(0.0, xx2 - xx1 + 1) * np.maximum(
0.0, yy2 - yy1 + 1)
iou = inter / (areas[i] + areas[order[1:]] - inter)
indexs = np.where(iou <= threshold)[0]
order = order[indexs + 1]
bbox = bbox_array[keep]
predicts_dict[object_name] = bbox.tolist()
predicts_dict = predicts_dict
return predicts_dict
def plot_result(self, src_img, predicts_dict, save_name=None):
"""
plot bounding boxes on source image.
Args:
src_img: source image
predicts_dict: {"stick": [[x1, y1, x2, y2, scores1], [...]]}.
"""
height_ratio = 1.0 * src_img.shape[0] / self._common_params[
"image_size"]
width_ratio = 1.0 * src_img.shape[1] / self._common_params["image_size"]
for object_name, bbox in predicts_dict.items():
for box in bbox:
xmin, ymin, xmax, ymax, score = box
src_xmin = xmin * width_ratio
src_ymin = ymin * height_ratio
src_xmax = xmax * width_ratio
src_ymax = ymax * height_ratio
score = float("%.3f" % score)
cv2.rectangle(src_img, (int(src_xmin), int(src_ymin)),
(int(src_xmax), int(src_ymax)), (0, 0, 255))
cv2.putText(src_img, object_name + str(score),
(int(src_xmin), int(src_ymin)), 1, 2, (0, 0, 255))
cv2.imshow("result", src_img)
if save_name is not None:
cv2.imwrite(save_name, src_img)
def object_track(self, predicts_dict, object_name="cup"):
"""track the specified object with maximum confidence.
Args:
object_name: object name.
"""
if self._motionProxy.getStiffnesses("Head") < 1.0:
self._motionProxy.setStiffnesses("Head", 1.0)
if self._motionProxy.getStiffnesses("LArm") < 1.0:
self._motionProxy.setStiffnesses("LArm", 1.0)
img_size = self._common_params["image_size"]
img_center_x = img_size / 2
img_center_y = img_size / 2
if predicts_dict.has_key(object_name):
predict_coords = predicts_dict[object_name]
predict_coords.sort(key=lambda coord: coord[-1], reverse=True)
predict_coord = predict_coords[0]
xmin, ymin, xmax, ymax, _ = predict_coord
center_x = (xmin + xmax) / 2
center_y = (ymin + ymax) / 2
angle_yaw = (center_x -
img_center_x) / (img_size) * self._cameraYawRange
angle_pitch = (center_y -
img_center_y) / (img_size) * self._cameraPitchRange
self._motionProxy.angleInterpolation(
["HeadPitch", "HeadYaw"],
[0.8 * angle_pitch, -0.8 * angle_yaw], 0.5, False)
head_pitch, head_yaw = self._motionProxy.getAngles("Head", False)
arm_angle = [
head_yaw - 7 / 180 * np.pi, head_pitch, -1.15, -0.035, -1.54,
0.01
]
self._motionProxy.setAngles("LArm", arm_angle, 0.2)
self._motionProxy.openHand("LHand")
xmax = xmin + w
ymax = ymin + h
#输出左上角和右下角坐标
return xmin, ymin, xmax, ymax, class_num, confidence
common_params = {'image_size': 448, 'num_classes': 20, 'batch_size': 1}
net_params = {
'cell_size': 7,
'boxes_per_cell': 2,
'weight_decay': 0.0005
} #网络结构参数
net = YoloTinyNet(common_params, net_params, test=True) #定义网络 传入字典 获得网络类对象
image = tf.placeholder(tf.float32, (1, 448, 448, 3)) # 图片大小
predicts = net.inference(image)
# print(predicts.shape)
sess = tf.Session()
np_img = cv2.imread('dog.jpg')
resized_img = cv2.resize(np_img, (448, 448))
np_img = cv2.cvtColor(resized_img, cv2.COLOR_BGR2RGB) #颜色通道转换
np_img = np_img.astype(np.float32)
np_img = np_img / 255.0 * 2 - 1 #预处理[-1,1]
np_img = np.reshape(np_img, (1, 448, 448, 3))
class ObjectDetection(VisualBasis):
def __init__(self, IP, cameraId=vd.kTopCamera, resolution=vd.kVGA):
super(ObjectDetection, self).__init__(IP, cameraId, resolution)
self._boundRect = []
self._cropKeep = 1
self._stickAngle = None # rad
#self._classes_name = ["stick"]
self._common_params = {'image_size': 448, 'num_classes': 1,
'batch_size':1}
self._net_params = {'cell_size': 7, 'boxes_per_cell':2, 'weight_decay': 0.0005}
self._net = YoloTinyNet(self._common_params, self._net_params, test=True)
#self._modelFile = "/home/meringue/Documents/python-nao-golf/yoloNet/models/train/model.ckpt-95000"
#self._objectRect = [0, 0, 0, 0]
self._objectName = None
def predict_single_object(self, image):
predicts = self._net.inference(image)
return predicts
def process_predicts(self, predicts):
p_classes = predicts[0, :, :, 0:1]
C = predicts[0, :, :, 1:3]
coordinate = predicts[0, :, :, 3:]
p_classes = np.reshape(p_classes, (7, 7, 1, 1))
C = np.reshape(C, (7, 7, 2, 1))
P = C * p_classes
index = np.argmax(P)
print("confidence = ", np.max(P))
index = np.unravel_index(index, P.shape)
class_num = index[3]
coordinate = np.reshape(coordinate, (7, 7, 2, 4))
max_coordinate = coordinate[index[0], index[1], index[2], :]
xcenter = max_coordinate[0]
ycenter = max_coordinate[1]
w = max_coordinate[2]
h = max_coordinate[3]
xcenter = (index[1] + xcenter) * (448/7.0)
ycenter = (index[0] + ycenter) * (448/7.0)
w = w * 448
h = h * 448
xmin = xcenter - w/2.0
ymin = ycenter - h/2.0
xmax = xmin + w
ymax = ymin + h
return [xmin, ymin, xmax, ymax], class_num
def showDetectResult(self, frame, rect, object_name):
object_min_xy = (int(rect[0]), int(rect[1]))
object_max_xy = (int(rect[2]), int(rect[3]))
cv2.rectangle(frame, object_min_xy, object_max_xy, (0, 0, 255))
cv2.putText(frame, object_name, object_min_xy, 2, 2, (0, 0, 255))
cv2.imshow("detect result", frame)
#cv2.waitKey(10)
