The following networks have been tested and known to work:
- AlexNet
- GoogLeNet v1 (also referred to as Inception-5h)
- ResNet-34/50/...
- SqueezeNet v1.1
- VGG-based FCN (semantical segmentation network)
- ENet (lightweight semantical segmentation network)
- VGG-based SSD (object detection network)
- MobileNet-based SSD (light-weight object detection network)
下面是我们将用到的一些函数。
在dnn中从磁盘加载图片：
- cv2.dnn.blobFromImage
- cv2.dnn.blobFromImages
用“create”方法直接从各种框架中导出模型：
- cv2.dnn.createCaffeImporter
- cv2.dnn.createTensorFlowImporter
- cv2.dnn.createTorchImporter
使用“读取”方法从磁盘直接加载序列化模型：
- cv2.dnn.readNetFromCaffe
- cv2.dnn.readNetFromTensorFlow
- cv2.dnn.readNetFromTorch
- cv2.dnn.readhTorchBlob
从磁盘加载完模型之后，可以用.forward方法来向前传播我们的图像，获取分类结果。
看样子就是好东西，那么，一起来安装：Installing OpenCV 3.3.0 on Ubuntu 16.04 LTS
You may meet the trouble in conflicting with python in anaconda3. Solve it as following:
lolo@lolo-UX303UB$ mv /usr/bin/python3
python3 python3.-config python3.4m-config python3m
python3. python3.4m python3-config python3m-config lolo: Move them away.
cmake -D CMAKE_BUILD_TYPE=RELEASE \
-D CMAKE_INSTALL_PREFIX=/usr/local/anaconda3 \
-D INSTALL_PYTHON_EXAMPLES=ON \
-D INSTALL_C_EXAMPLES=OFF \
-D OPENCV_EXTRA_MODULES_PATH=/home/unsw/Android/opencv-3.3./opencv_contrib-3.3./modules \
-D PYTHON_EXECUTABLE=/usr/local/anaconda3/bin/python3. \
-D BUILD_EXAMPLES=ON ..
Done :-)
Installing ref:
Now, you have got everything. Let's practice.
From: http://www.pyimagesearch.com/2017/09/11/object-detection-with-deep-learning-and-opencv/
In the first part of today’s post on object detection using deep learning we’ll discuss Single Shot Detectors and MobileNets.
SSD Paper: http://lib.csdn.net/article/deeplearning/53059
SSD Paper: https://arxiv.org/abs/1512.02325 [Origin]
When it comes to deep learning-based object detection there are three primary object detection methods that you’ll likely encounter:
- Faster R-CNNs (Girshick et al., 2015) 7 FPS
- You Only Look Once (YOLO) (Redmon and Farhadi, 2015) 155 FPS.
- Single Shot Detectors (SSDs) (Liu et al., 2015)
Code analysis:
# USAGE
# python deep_learning_object_detection.py --image images/example_01.jpg \
# --prototxt MobileNetSSD_deploy.prototxt.txt --model MobileNetSSD_deploy.caffemodel # import the necessary packages
import numpy as np
import argparse
import cv2 # construct the argument parse and parse the arguments
ap = argparse.ArgumentParser()
ap.add_argument("-i", "--image", required=True,
help="path to input image")
ap.add_argument("-p", "--prototxt", required=True,
help="path to Caffe 'deploy' prototxt file")
ap.add_argument("-m", "--model", required=True,
help="path to Caffe pre-trained model")
ap.add_argument("-c", "--confidence", type=float, default=0.2,
help="minimum probability to filter weak detections")
args = vars(ap.parse_args()) # initialize the list of class labels MobileNet SSD was trained to
# detect, then generate a set of bounding box colors for each class
CLASSES = ["background", "aeroplane", "bicycle", "bird", "boat",
"bottle", "bus", "car", "cat", "chair", "cow", "diningtable",
"dog", "horse", "motorbike", "person", "pottedplant", "sheep",
"sofa", "train", "tvmonitor"]
COLORS = np.random.uniform(0, 255, size=(len(CLASSES), 3)) # load our serialized model from disk
print("[INFO] loading model...")
net = cv2.dnn.readNetFromCaffe(args["prototxt"], args["model"]) # load the input image and construct an input blob for the image
# by resizing to a fixed 300x300 pixels and then normalizing it
# (note: normalization is done via the authors of the MobileNet SSD
# implementation)
image = cv2.imread(args["image"])
(h, w) = image.shape[:2]
blob = cv2.dnn.blobFromImage(image, 0.007843, (300, 300), 127.5) # --> NCHW # pass the blob through the network and obtain the detections and
# predictions
print("[INFO] computing object detections...")
net.setInput(blob)
detections = net.forward() # --> net.forward
# loop over the detections
for i in np.arange(0, detections.shape[2]):
# extract the confidence (i.e., probability) associated with the
# prediction
confidence = detections[0, 0, i, 2] # filter out weak detections by ensuring the `confidence` is
# greater than the minimum confidence
if confidence > args["confidence"]:
# extract the index of the class label from the `detections`,
# then compute the (x, y)-coordinates of the bounding box for
# the object
idx = int(detections[0, 0, i, 1])
box= detections[0, 0, i, 3:7] * np.array([w, h, w, h])
(startX, startY, endX, endY) = box.astype("int") # display the prediction
label = "{}: {:.2f}%".format(CLASSES[idx], confidence * 100)
print("[INFO] {}".format(label))
cv2.rectangle(image, (startX, startY), (endX, endY), COLORS[idx], 2)
y = startY - 15 if startY - 15 > 15 else startY + 15
cv2.putText(image, label, (startX, y), cv2.FONT_HERSHEY_SIMPLEX, 0.5, COLORS[idx], 2) # show the output image
cv2.imshow("Output", image)
cv2.waitKey(0)
NCHW
There is a comment that explains this, but in a different source file, ConvolutionalNodes.h, pasted below.
Note that the NVidia abbreviations refer to row-major layout, so to map them to column-major tensor indices are used by CNTK, you will need to reverse their order. E.g. cudnn stores images in “NCHW,” which is a [W x H x C x N] tensor in CNTK notation (W being the fastest-changing dimension; and there are N objects of dimension [W x H x C] concatenated).
Note that the “legacy” (non-cuDNN) memory layout is old code written before NCHW became the standard, so we are likely phasing out the old representation eventually.
net.forward
[INFO] loading model...
[INFO] computing object detections...
(1, 1, 2, 7)
[[[[ 0. 12. 0.95878285 0.49966827 0.6235761 0.69597626 0.87614471]
[ 0. 15. 0.99952459 0.04266162 0.20033446 0.45632178 0.84977102]]]]
[INFO] dog: 95.88%
[INFO] person: 99.95%