Upload 9 files

.gitattributes

@@ -32,3 +32,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+age_net.caffemodel filter=lfs diff=lfs merge=lfs -text
+gender_net.caffemodel filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,3 +1,110 @@
----
-license: c-uda
----
+# Gender-and-Age-Detection   <img alt="GitHub" src="https://img.shields.io/github/license/smahesh29/Gender-and-Age-Detection">
+
+
+<h2>Objective :</h2>
+<p>To build a gender and age detector that can approximately guess the gender and age of the person (face) in a picture or through webcam.</p>
+
+<h2>About the Project :</h2>
+<p>In this Python Project, I had used Deep Learning to accurately identify the gender and age of a person from a single image of a face. I used the models trained by <a href="https://talhassner.github.io/home/projects/Adience/Adience-data.html">Tal Hassner and Gil Levi</a>. The predicted gender may be one of ‘Male’ and ‘Female’, and the predicted age may be one of the following ranges- (0 – 2), (4 – 6), (8 – 12), (15 – 20), (25 – 32), (38 – 43), (48 – 53), (60 – 100) (8 nodes in the final softmax layer). It is very difficult to accurately guess an exact age from a single image because of factors like makeup, lighting, obstructions, and facial expressions. And so, I made this a classification problem instead of making it one of regression.</p>
+
+<h2>Dataset :</h2>
+<p>For this python project, I had used the Adience dataset; the dataset is available in the public domain and you can find it <a href="https://www.kaggle.com/ttungl/adience-benchmark-gender-and-age-classification">here</a>. This dataset serves as a benchmark for face photos and is inclusive of various real-world imaging conditions like noise, lighting, pose, and appearance. The images have been collected from Flickr albums and distributed under the Creative Commons (CC) license. It has a total of 26,580 photos of 2,284 subjects in eight age ranges (as mentioned above) and is about 1GB in size. The models I used had been trained on this dataset.</p>
+
+<h2>Additional Python Libraries Required :</h2>
+<ul>
+  <li>OpenCV</li>
+  
+       pip install opencv-python
+</ul>
+<ul>
+ <li>argparse</li>
+  
+       pip install argparse
+</ul>
+
+<h2>The contents of this Project :</h2>
+<ul>
+  <li>opencv_face_detector.pbtxt</li>
+  <li>opencv_face_detector_uint8.pb</li>
+  <li>age_deploy.prototxt</li>
+  <li>age_net.caffemodel</li>
+  <li>gender_deploy.prototxt</li>
+  <li>gender_net.caffemodel</li>
+  <li>a few pictures to try the project on</li>
+  <li>detect.py</li>
+ </ul>
+ <p>For face detection, we have a .pb file- this is a protobuf file (protocol buffer); it holds the graph definition and the trained weights of the model. We can use this to run the trained model. And while a .pb file holds the protobuf in binary format, one with the .pbtxt extension holds it in text format. These are TensorFlow files. For age and gender, the .prototxt files describe the network configuration and the .caffemodel file defines the internal states of the parameters of the layers.</p>
+ 
+ <h2>Usage :</h2>
+ <ul>
+  <li>Download my Repository</li>
+  <li>Open your Command Prompt or Terminal and change directory to the folder where all the files are present.</li>
+  <li><b>Detecting Gender and Age of face in Image</b> Use Command :</li>
+  
+      python detect.py --image <image_name>
+</ul>
+  <p><b>Note: </b>The Image should be present in same folder where all the files are present</p> 
+<ul>
+  <li><b>Detecting Gender and Age of face through webcam</b> Use Command :</li>
+  
+      python detect.py
+</ul>
+<ul>
+  <li>Press <b>Ctrl + C</b> to stop the program execution.</li>
+</ul>
+
+# Working:
+[![Watch the video](https://img.youtube.com/vi/ReeccRD21EU/0.jpg)](https://youtu.be/ReeccRD21EU)
+
+<h2>Examples :</h2>
+<p><b>NOTE:- I downloaded the images from Google,if you have any query or problem i can remove them, i just used it for Educational purpose.</b></p>
+
+    >python detect.py --image girl1.jpg
+    Gender: Female
+    Age: 25-32 years
+    
+<img src="Example/Detecting age and gender girl1.png">
+
+    >python detect.py --image girl2.jpg
+    Gender: Female
+    Age: 8-12 years
+    
+<img src="Example/Detecting age and gender girl2.png">
+
+    >python detect.py --image kid1.jpg
+    Gender: Male
+    Age: 4-6 years    
+    
+<img src="Example/Detecting age and gender kid1.png">
+
+    >python detect.py --image kid2.jpg
+    Gender: Female
+    Age: 4-6 years  
+    
+<img src="Example/Detecting age and gender kid2.png">
+
+    >python detect.py --image man1.jpg
+    Gender: Male
+    Age: 38-43 years
+    
+<img src="Example/Detecting age and gender man1.png">
+
+    >python detect.py --image man2.jpg
+    Gender: Male
+    Age: 25-32 years
+    
+<img src="Example/Detecting age and gender man2.png">
+
+    >python detect.py --image woman1.jpg
+    Gender: Female
+    Age: 38-43 years
+    
+<img src="Example/Detecting age and gender woman1.png">
+    
+# Support :
+If you found this project helpful or you learned something from the source code and want to thank me, consider me to pay my internet bills. This would encourage me to create many such projects 👨🏻‍💻
+<ul>
+    <li><a href="https://www.paypal.me/smahesh29"><b>PayPal</b></a></li>
+    <li><a href="https://imjo.in/XNZDCJ"><b>₹ (INR)</b></a></li>
+    <li><b>UPI ID :</b> maheshusa29@oksbi</li>
+</ul>           

_config.yml

@@ -0,0 +1 @@
+theme: jekyll-theme-tactile

age_deploy.prototxt

@@ -0,0 +1,175 @@
+name: "CaffeNet"
+input: "data"
+input_dim: 1
+input_dim: 3
+input_dim: 227
+input_dim: 227
+layers {
+  name: "conv1"
+  type: CONVOLUTION
+  bottom: "data"
+  top: "conv1"
+  convolution_param {
+    num_output: 96
+    kernel_size: 7
+    stride: 4
+  }
+}
+layers {
+  name: "relu1"
+  type: RELU
+  bottom: "conv1"
+  top: "conv1"
+}
+layers {
+  name: "pool1"
+  type: POOLING
+  bottom: "conv1"
+  top: "pool1"
+  pooling_param {
+    pool: MAX
+    kernel_size: 3
+    stride: 2
+  }
+}
+layers {
+  name: "norm1"
+  type: LRN
+  bottom: "pool1"
+  top: "norm1"
+  lrn_param {
+    local_size: 5
+    alpha: 0.0001
+    beta: 0.75
+  }
+}
+layers {
+  name: "conv2"
+  type: CONVOLUTION
+  bottom: "norm1"
+  top: "conv2"
+  convolution_param {
+    num_output: 256
+    pad: 2
+    kernel_size: 5
+  }
+}
+layers {
+  name: "relu2"
+  type: RELU
+  bottom: "conv2"
+  top: "conv2"
+}
+layers {
+  name: "pool2"
+  type: POOLING
+  bottom: "conv2"
+  top: "pool2"
+  pooling_param {
+    pool: MAX
+    kernel_size: 3
+    stride: 2
+  }
+}
+layers {
+  name: "norm2"
+  type: LRN
+  bottom: "pool2"
+  top: "norm2"
+  lrn_param {
+    local_size: 5
+    alpha: 0.0001
+    beta: 0.75
+  }
+}
+layers {
+  name: "conv3"
+  type: CONVOLUTION
+  bottom: "norm2"
+  top: "conv3"
+  convolution_param {
+    num_output: 384
+    pad: 1
+    kernel_size: 3
+  }
+}
+layers{
+  name: "relu3" 
+  type: RELU
+  bottom: "conv3"
+  top: "conv3"
+}
+layers {
+  name: "pool5"
+  type: POOLING
+  bottom: "conv3"
+  top: "pool5"
+  pooling_param {
+    pool: MAX
+    kernel_size: 3
+    stride: 2
+  }
+}
+layers {
+  name: "fc6"
+  type: INNER_PRODUCT
+  bottom: "pool5"
+  top: "fc6"
+  inner_product_param {
+    num_output: 512
+  }
+}
+layers {
+  name: "relu6"
+  type: RELU
+  bottom: "fc6"
+  top: "fc6"
+}
+layers {
+  name: "drop6"
+  type: DROPOUT
+  bottom: "fc6"
+  top: "fc6"
+  dropout_param {
+    dropout_ratio: 0.5
+  }
+}
+layers {
+  name: "fc7"
+  type: INNER_PRODUCT
+  bottom: "fc6"
+  top: "fc7"
+  inner_product_param {
+    num_output: 512
+  }
+}
+layers {
+  name: "relu7"
+  type: RELU
+  bottom: "fc7"
+  top: "fc7"
+}
+layers {
+  name: "drop7"
+  type: DROPOUT
+  bottom: "fc7"
+  top: "fc7"
+  dropout_param {
+    dropout_ratio: 0.5
+  }
+}
+layers {
+  name: "fc8"
+  type: INNER_PRODUCT
+  bottom: "fc7"
+  top: "fc8"
+  inner_product_param {
+    num_output: 8
+  }
+}
+layers {
+  name: "prob"
+  type: SOFTMAX
+  bottom: "fc8"
+  top: "prob"
+}

age_net.caffemodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6dde5d07df5ca1d66ff39e525693f05ccfb9d2c437e188fdd1a10d42e57fabd6
+size 45661480

detect.py

@@ -0,0 +1,127 @@
+# A Gender and Age Detection program by Mahesh Sawant
+import os
+import pandas as pd
+import cv2
+import math
+import argparse
+
+dic = {"images": [], "gender": [], "age": []}
+
+
+def highlightFace(net, frame, conf_threshold=0.7):
+    frameOpencvDnn = frame.copy()
+    frameHeight = frameOpencvDnn.shape[0]
+    frameWidth = frameOpencvDnn.shape[1]
+    blob = cv2.dnn.blobFromImage(frameOpencvDnn, 1.0, (300, 300), [104, 117, 123], True, False)
+
+    net.setInput(blob)
+    detections = net.forward()
+    faceBoxes = []
+    for i in range(detections.shape[2]):
+        confidence = detections[0, 0, i, 2]
+        if confidence > conf_threshold:
+            x1 = int(detections[0, 0, i, 3] * frameWidth)
+            y1 = int(detections[0, 0, i, 4] * frameHeight)
+            x2 = int(detections[0, 0, i, 5] * frameWidth)
+            y2 = int(detections[0, 0, i, 6] * frameHeight)
+            faceBoxes.append([x1, y1, x2, y2])
+            cv2.rectangle(frameOpencvDnn, (x1, y1), (x2, y2), (0, 255, 0), int(round(frameHeight / 150)), 8)
+    return frameOpencvDnn, faceBoxes
+
+
+def process_image(image):
+    # parser=argparse.ArgumentParser()
+    # parser.add_argument('--image')
+    #
+    # args=parser.parse_args()
+
+    faceProto = "opencv_face_detector.pbtxt"
+    faceModel = "opencv_face_detector_uint8.pb"
+    ageProto = "age_deploy.prototxt"
+    ageModel = "age_net.caffemodel"
+    genderProto = "gender_deploy.prototxt"
+    genderModel = "gender_net.caffemodel"
+
+    MODEL_MEAN_VALUES = (78.4263377603, 87.7689143744, 114.895847746)
+    ageList = ['(0-2)', '(4-6)', '(8-12)', '(15-20)', '(25-32)', '(38-43)', '(48-53)', '(60-100)']
+    genderList = ['Male', 'Female']
+
+    faceNet = cv2.dnn.readNet(faceModel, faceProto)
+    ageNet = cv2.dnn.readNet(ageModel, ageProto)
+    genderNet = cv2.dnn.readNet(genderModel, genderProto)
+
+    video = cv2.VideoCapture(image)
+    padding = 20
+    while cv2.waitKey(1) < 0:
+        try:
+            hasFrame, frame = video.read()
+            if not hasFrame:
+                cv2.waitKey()
+                break
+
+            resultImg, faceBoxes = highlightFace(faceNet, frame)
+            if not faceBoxes:
+                print("No face detected")
+
+            for faceBox in faceBoxes:
+                face = frame[max(0, faceBox[1] - padding):
+                             min(faceBox[3] + padding, frame.shape[0] - 1), max(0, faceBox[0] - padding)
+                                                                            :min(faceBox[2] + padding,
+                                                                                 frame.shape[1] - 1)]
+
+                blob = cv2.dnn.blobFromImage(face, 1.0, (227, 227), MODEL_MEAN_VALUES, swapRB=False)
+                genderNet.setInput(blob)
+                genderPreds = genderNet.forward()
+                gender = genderList[genderPreds[0].argmax()]
+                print(f'Gender: {gender}')
+
+                ageNet.setInput(blob)
+                agePreds = ageNet.forward()
+                age = ageList[agePreds[0].argmax()]
+
+                print(f'Age: {age[1:-1]} years')
+                dic['images'].append(image)
+                dic['gender'].append(gender)
+                dic['age'].append(age[1:-1])
+
+                # cv2.putText(resultImg, f'{gender}, {age}', (faceBox[0], faceBox[1]-10), cv2.FONT_HERSHEY_SIMPLEX, 0.8, (0,255,255), 2, cv2.LINE_AA)
+                # cv2.imshow("Detecting age and gender", resultImg)
+        except Exception as e:
+            continue
+
+
+
+import boto3
+s3 = boto3.resource(
+    service_name = 's3',
+    region_name = 'ap-south-1',
+    aws_access_key_id = 'AKIAYNE4X3VIWUPXM75R',
+    aws_secret_access_key ='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'
+) 
+def download_s3_folder(bucket, folder, local_dir='./images'):
+    bucket = s3.Bucket(bucket)
+    for obj in bucket.objects.filter(Prefix=folder):
+        target = obj.key if local_dir is None \
+            else os.path.join(local_dir, os.path.relpath(obj.key, folder))
+        if not os.path.exists(os.path.dirname(target)):
+            os.makedirs(os.path.dirname(target))
+        if obj.key[-1] == '/':
+            continue
+        bucket.download_file(obj.key, target)
+
+
+def predict_age_gender():
+    image = os.listdir('images')
+
+    for img in image:
+        img = './images/' + img
+        process_image(img)
+    print(dic)
+    df = pd.DataFrame.from_dict(dic, orient='index').transpose()
+    df.head()
+    df.to_excel("./output/result_s3.xls")
+
+download_s3_folder('genderagedata','input_images')
+predict_age_gender()
+
+s3.Bucket('genderagedata').upload_file(Filename='./output/result_s3.xls', Key='output_images/result.xls')

gender_deploy.prototxt

@@ -0,0 +1,175 @@
+name: "CaffeNet"
+input: "data"
+input_dim: 10
+input_dim: 3
+input_dim: 227
+input_dim: 227
+layers {
+  name: "conv1"
+  type: CONVOLUTION
+  bottom: "data"
+  top: "conv1"
+  convolution_param {
+    num_output: 96
+    kernel_size: 7
+    stride: 4
+  }
+}
+layers {
+  name: "relu1"
+  type: RELU
+  bottom: "conv1"
+  top: "conv1"
+}
+layers {
+  name: "pool1"
+  type: POOLING
+  bottom: "conv1"
+  top: "pool1"
+  pooling_param {
+    pool: MAX
+    kernel_size: 3
+    stride: 2
+  }
+}
+layers {
+  name: "norm1"
+  type: LRN
+  bottom: "pool1"
+  top: "norm1"
+  lrn_param {
+    local_size: 5
+    alpha: 0.0001
+    beta: 0.75
+  }
+}
+layers {
+  name: "conv2"
+  type: CONVOLUTION
+  bottom: "norm1"
+  top: "conv2"
+  convolution_param {
+    num_output: 256
+    pad: 2
+    kernel_size: 5
+  }
+}
+layers {
+  name: "relu2"
+  type: RELU
+  bottom: "conv2"
+  top: "conv2"
+}
+layers {
+  name: "pool2"
+  type: POOLING
+  bottom: "conv2"
+  top: "pool2"
+  pooling_param {
+    pool: MAX
+    kernel_size: 3
+    stride: 2
+  }
+}
+layers {
+  name: "norm2"
+  type: LRN
+  bottom: "pool2"
+  top: "norm2"
+  lrn_param {
+    local_size: 5
+    alpha: 0.0001
+    beta: 0.75
+  }
+}
+layers {
+  name: "conv3"
+  type: CONVOLUTION
+  bottom: "norm2"
+  top: "conv3"
+  convolution_param {
+    num_output: 384
+    pad: 1
+    kernel_size: 3
+  }
+}
+layers{
+  name: "relu3" 
+  type: RELU
+  bottom: "conv3"
+  top: "conv3"
+}
+layers {
+  name: "pool5"
+  type: POOLING
+  bottom: "conv3"
+  top: "pool5"
+  pooling_param {
+    pool: MAX
+    kernel_size: 3
+    stride: 2
+  }
+}
+layers {
+  name: "fc6"
+  type: INNER_PRODUCT
+  bottom: "pool5"
+  top: "fc6"
+  inner_product_param {
+    num_output: 512
+  }
+}
+layers {
+  name: "relu6"
+  type: RELU
+  bottom: "fc6"
+  top: "fc6"
+}
+layers {
+  name: "drop6"
+  type: DROPOUT
+  bottom: "fc6"
+  top: "fc6"
+  dropout_param {
+    dropout_ratio: 0.5
+  }
+}
+layers {
+  name: "fc7"
+  type: INNER_PRODUCT
+  bottom: "fc6"
+  top: "fc7"
+  inner_product_param {
+    num_output: 512
+  }
+}
+layers {
+  name: "relu7"
+  type: RELU
+  bottom: "fc7"
+  top: "fc7"
+}
+layers {
+  name: "drop7"
+  type: DROPOUT
+  bottom: "fc7"
+  top: "fc7"
+  dropout_param {
+    dropout_ratio: 0.5
+  }
+}
+layers {
+  name: "fc8"
+  type: INNER_PRODUCT
+  bottom: "fc7"
+  top: "fc8"
+  inner_product_param {
+    num_output: 2
+  }
+}
+layers {
+  name: "prob"
+  type: SOFTMAX
+  bottom: "fc8"
+  top: "prob"
+}

gender_net.caffemodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac7571b281ae078817764b645a20541bd6aa1babeac20a45e6d8de7d61ba0e50
+size 45649168

opencv_face_detector.pbtxt

@@ -0,0 +1,2362 @@
+node {
+  name: "data"
+  op: "Placeholder"
+  attr {
+    key: "dtype"
+    value {
+      type: DT_FLOAT
+    }
+  }
+}
+node {
+  name: "data_bn/FusedBatchNorm"
+  op: "FusedBatchNorm"
+  input: "data:0"
+  input: "data_bn/gamma"
+  input: "data_bn/beta"
+  input: "data_bn/mean"
+  input: "data_bn/std"
+  attr {
+    key: "epsilon"
+    value {
+      f: 1.00099996416e-05
+    }
+  }
+}
+node {
+  name: "data_scale/Mul"
+  op: "Mul"
+  input: "data_bn/FusedBatchNorm"
+  input: "data_scale/mul"
+}
+node {
+  name: "data_scale/BiasAdd"
+  op: "BiasAdd"
+  input: "data_scale/Mul"
+  input: "data_scale/add"
+}
+node {
+  name: "SpaceToBatchND/block_shape"
+  op: "Const"
+  attr {
+    key: "value"
+    value {
+      tensor {
+        dtype: DT_INT32
+        tensor_shape {
+          dim {
+            size: 2
+          }
+        }
+        int_val: 1
+        int_val: 1
+      }
+    }
+  }
+}
+node {
+  name: "SpaceToBatchND/paddings"
+  op: "Const"
+  attr {
+    key: "value"
+    value {
+      tensor {
+        dtype: DT_INT32
+        tensor_shape {
+          dim {
+            size: 2
+          }
+          dim {
+            size: 2
+          }
+        }
+        int_val: 3
+        int_val: 3
+        int_val: 3
+        int_val: 3
+      }
+    }
+  }
+}
+node {
+  name: "Pad"
+  op: "SpaceToBatchND"
+  input: "data_scale/BiasAdd"
+  input: "SpaceToBatchND/block_shape"
+  input: "SpaceToBatchND/paddings"
+}
+node {
+  name: "conv1_h/Conv2D"
+  op: "Conv2D"
+  input: "Pad"
+  input: "conv1_h/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "VALID"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 2
+        i: 2
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv1_h/BiasAdd"
+  op: "BiasAdd"
+  input: "conv1_h/Conv2D"
+  input: "conv1_h/bias"
+}
+node {
+  name: "BatchToSpaceND"
+  op: "BatchToSpaceND"
+  input: "conv1_h/BiasAdd"
+}
+node {
+  name: "conv1_bn_h/FusedBatchNorm"
+  op: "FusedBatchNorm"
+  input: "BatchToSpaceND"
+  input: "conv1_bn_h/gamma"
+  input: "conv1_bn_h/beta"
+  input: "conv1_bn_h/mean"
+  input: "conv1_bn_h/std"
+  attr {
+    key: "epsilon"
+    value {
+      f: 1.00099996416e-05
+    }
+  }
+}
+node {
+  name: "conv1_scale_h/Mul"
+  op: "Mul"
+  input: "conv1_bn_h/FusedBatchNorm"
+  input: "conv1_scale_h/mul"
+}
+node {
+  name: "conv1_scale_h/BiasAdd"
+  op: "BiasAdd"
+  input: "conv1_scale_h/Mul"
+  input: "conv1_scale_h/add"
+}
+node {
+  name: "Relu"
+  op: "Relu"
+  input: "conv1_scale_h/BiasAdd"
+}
+node {
+  name: "conv1_pool/MaxPool"
+  op: "MaxPool"
+  input: "Relu"
+  attr {
+    key: "ksize"
+    value {
+      list {
+        i: 1
+        i: 3
+        i: 3
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 2
+        i: 2
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "layer_64_1_conv1_h/Conv2D"
+  op: "Conv2D"
+  input: "conv1_pool/MaxPool"
+  input: "layer_64_1_conv1_h/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "layer_64_1_bn2_h/FusedBatchNorm"
+  op: "BiasAdd"
+  input: "layer_64_1_conv1_h/Conv2D"
+  input: "layer_64_1_conv1_h/Conv2D_bn_offset"
+}
+node {
+  name: "layer_64_1_scale2_h/Mul"
+  op: "Mul"
+  input: "layer_64_1_bn2_h/FusedBatchNorm"
+  input: "layer_64_1_scale2_h/mul"
+}
+node {
+  name: "layer_64_1_scale2_h/BiasAdd"
+  op: "BiasAdd"
+  input: "layer_64_1_scale2_h/Mul"
+  input: "layer_64_1_scale2_h/add"
+}
+node {
+  name: "Relu_1"
+  op: "Relu"
+  input: "layer_64_1_scale2_h/BiasAdd"
+}
+node {
+  name: "layer_64_1_conv2_h/Conv2D"
+  op: "Conv2D"
+  input: "Relu_1"
+  input: "layer_64_1_conv2_h/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "add"
+  op: "Add"
+  input: "layer_64_1_conv2_h/Conv2D"
+  input: "conv1_pool/MaxPool"
+}
+node {
+  name: "layer_128_1_bn1_h/FusedBatchNorm"
+  op: "FusedBatchNorm"
+  input: "add"
+  input: "layer_128_1_bn1_h/gamma"
+  input: "layer_128_1_bn1_h/beta"
+  input: "layer_128_1_bn1_h/mean"
+  input: "layer_128_1_bn1_h/std"
+  attr {
+    key: "epsilon"
+    value {
+      f: 1.00099996416e-05
+    }
+  }
+}
+node {
+  name: "layer_128_1_scale1_h/Mul"
+  op: "Mul"
+  input: "layer_128_1_bn1_h/FusedBatchNorm"
+  input: "layer_128_1_scale1_h/mul"
+}
+node {
+  name: "layer_128_1_scale1_h/BiasAdd"
+  op: "BiasAdd"
+  input: "layer_128_1_scale1_h/Mul"
+  input: "layer_128_1_scale1_h/add"
+}
+node {
+  name: "Relu_2"
+  op: "Relu"
+  input: "layer_128_1_scale1_h/BiasAdd"
+}
+node {
+  name: "layer_128_1_conv_expand_h/Conv2D"
+  op: "Conv2D"
+  input: "Relu_2"
+  input: "layer_128_1_conv_expand_h/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 2
+        i: 2
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "layer_128_1_conv1_h/Conv2D"
+  op: "Conv2D"
+  input: "Relu_2"
+  input: "layer_128_1_conv1_h/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 2
+        i: 2
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "layer_128_1_bn2/FusedBatchNorm"
+  op: "BiasAdd"
+  input: "layer_128_1_conv1_h/Conv2D"
+  input: "layer_128_1_conv1_h/Conv2D_bn_offset"
+}
+node {
+  name: "layer_128_1_scale2/Mul"
+  op: "Mul"
+  input: "layer_128_1_bn2/FusedBatchNorm"
+  input: "layer_128_1_scale2/mul"
+}
+node {
+  name: "layer_128_1_scale2/BiasAdd"
+  op: "BiasAdd"
+  input: "layer_128_1_scale2/Mul"
+  input: "layer_128_1_scale2/add"
+}
+node {
+  name: "Relu_3"
+  op: "Relu"
+  input: "layer_128_1_scale2/BiasAdd"
+}
+node {
+  name: "layer_128_1_conv2/Conv2D"
+  op: "Conv2D"
+  input: "Relu_3"
+  input: "layer_128_1_conv2/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "add_1"
+  op: "Add"
+  input: "layer_128_1_conv2/Conv2D"
+  input: "layer_128_1_conv_expand_h/Conv2D"
+}
+node {
+  name: "layer_256_1_bn1/FusedBatchNorm"
+  op: "FusedBatchNorm"
+  input: "add_1"
+  input: "layer_256_1_bn1/gamma"
+  input: "layer_256_1_bn1/beta"
+  input: "layer_256_1_bn1/mean"
+  input: "layer_256_1_bn1/std"
+  attr {
+    key: "epsilon"
+    value {
+      f: 1.00099996416e-05
+    }
+  }
+}
+node {
+  name: "layer_256_1_scale1/Mul"
+  op: "Mul"
+  input: "layer_256_1_bn1/FusedBatchNorm"
+  input: "layer_256_1_scale1/mul"
+}
+node {
+  name: "layer_256_1_scale1/BiasAdd"
+  op: "BiasAdd"
+  input: "layer_256_1_scale1/Mul"
+  input: "layer_256_1_scale1/add"
+}
+node {
+  name: "Relu_4"
+  op: "Relu"
+  input: "layer_256_1_scale1/BiasAdd"
+}
+node {
+  name: "SpaceToBatchND_1/paddings"
+  op: "Const"
+  attr {
+    key: "value"
+    value {
+      tensor {
+        dtype: DT_INT32
+        tensor_shape {
+          dim {
+            size: 2
+          }
+          dim {
+            size: 2
+          }
+        }
+        int_val: 1
+        int_val: 1
+        int_val: 1
+        int_val: 1
+      }
+    }
+  }
+}
+node {
+  name: "layer_256_1_conv_expand/Conv2D"
+  op: "Conv2D"
+  input: "Relu_4"
+  input: "layer_256_1_conv_expand/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 2
+        i: 2
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv4_3_norm/l2_normalize"
+  op: "L2Normalize"
+  input: "Relu_4:0"
+  input: "conv4_3_norm/l2_normalize/Sum/reduction_indices"
+}
+node {
+  name: "conv4_3_norm/mul_1"
+  op: "Mul"
+  input: "conv4_3_norm/l2_normalize"
+  input: "conv4_3_norm/mul"
+}
+node {
+  name: "conv4_3_norm_mbox_loc/Conv2D"
+  op: "Conv2D"
+  input: "conv4_3_norm/mul_1"
+  input: "conv4_3_norm_mbox_loc/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv4_3_norm_mbox_loc/BiasAdd"
+  op: "BiasAdd"
+  input: "conv4_3_norm_mbox_loc/Conv2D"
+  input: "conv4_3_norm_mbox_loc/bias"
+}
+node {
+  name: "flatten/Reshape"
+  op: "Flatten"
+  input: "conv4_3_norm_mbox_loc/BiasAdd"
+}
+node {
+  name: "conv4_3_norm_mbox_conf/Conv2D"
+  op: "Conv2D"
+  input: "conv4_3_norm/mul_1"
+  input: "conv4_3_norm_mbox_conf/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv4_3_norm_mbox_conf/BiasAdd"
+  op: "BiasAdd"
+  input: "conv4_3_norm_mbox_conf/Conv2D"
+  input: "conv4_3_norm_mbox_conf/bias"
+}
+node {
+  name: "flatten_6/Reshape"
+  op: "Flatten"
+  input: "conv4_3_norm_mbox_conf/BiasAdd"
+}
+node {
+  name: "Pad_1"
+  op: "SpaceToBatchND"
+  input: "Relu_4"
+  input: "SpaceToBatchND/block_shape"
+  input: "SpaceToBatchND_1/paddings"
+}
+node {
+  name: "layer_256_1_conv1/Conv2D"
+  op: "Conv2D"
+  input: "Pad_1"
+  input: "layer_256_1_conv1/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "VALID"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 2
+        i: 2
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "layer_256_1_bn2/FusedBatchNorm"
+  op: "BiasAdd"
+  input: "layer_256_1_conv1/Conv2D"
+  input: "layer_256_1_conv1/Conv2D_bn_offset"
+}
+node {
+  name: "BatchToSpaceND_1"
+  op: "BatchToSpaceND"
+  input: "layer_256_1_bn2/FusedBatchNorm"
+}
+node {
+  name: "layer_256_1_scale2/Mul"
+  op: "Mul"
+  input: "BatchToSpaceND_1"
+  input: "layer_256_1_scale2/mul"
+}
+node {
+  name: "layer_256_1_scale2/BiasAdd"
+  op: "BiasAdd"
+  input: "layer_256_1_scale2/Mul"
+  input: "layer_256_1_scale2/add"
+}
+node {
+  name: "Relu_5"
+  op: "Relu"
+  input: "layer_256_1_scale2/BiasAdd"
+}
+node {
+  name: "layer_256_1_conv2/Conv2D"
+  op: "Conv2D"
+  input: "Relu_5"
+  input: "layer_256_1_conv2/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "add_2"
+  op: "Add"
+  input: "layer_256_1_conv2/Conv2D"
+  input: "layer_256_1_conv_expand/Conv2D"
+}
+node {
+  name: "layer_512_1_bn1/FusedBatchNorm"
+  op: "FusedBatchNorm"
+  input: "add_2"
+  input: "layer_512_1_bn1/gamma"
+  input: "layer_512_1_bn1/beta"
+  input: "layer_512_1_bn1/mean"
+  input: "layer_512_1_bn1/std"
+  attr {
+    key: "epsilon"
+    value {
+      f: 1.00099996416e-05
+    }
+  }
+}
+node {
+  name: "layer_512_1_scale1/Mul"
+  op: "Mul"
+  input: "layer_512_1_bn1/FusedBatchNorm"
+  input: "layer_512_1_scale1/mul"
+}
+node {
+  name: "layer_512_1_scale1/BiasAdd"
+  op: "BiasAdd"
+  input: "layer_512_1_scale1/Mul"
+  input: "layer_512_1_scale1/add"
+}
+node {
+  name: "Relu_6"
+  op: "Relu"
+  input: "layer_512_1_scale1/BiasAdd"
+}
+node {
+  name: "layer_512_1_conv_expand_h/Conv2D"
+  op: "Conv2D"
+  input: "Relu_6"
+  input: "layer_512_1_conv_expand_h/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "layer_512_1_conv1_h/Conv2D"
+  op: "Conv2D"
+  input: "Relu_6"
+  input: "layer_512_1_conv1_h/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "layer_512_1_bn2_h/FusedBatchNorm"
+  op: "BiasAdd"
+  input: "layer_512_1_conv1_h/Conv2D"
+  input: "layer_512_1_conv1_h/Conv2D_bn_offset"
+}
+node {
+  name: "layer_512_1_scale2_h/Mul"
+  op: "Mul"
+  input: "layer_512_1_bn2_h/FusedBatchNorm"
+  input: "layer_512_1_scale2_h/mul"
+}
+node {
+  name: "layer_512_1_scale2_h/BiasAdd"
+  op: "BiasAdd"
+  input: "layer_512_1_scale2_h/Mul"
+  input: "layer_512_1_scale2_h/add"
+}
+node {
+  name: "Relu_7"
+  op: "Relu"
+  input: "layer_512_1_scale2_h/BiasAdd"
+}
+node {
+  name: "layer_512_1_conv2_h/convolution/SpaceToBatchND"
+  op: "SpaceToBatchND"
+  input: "Relu_7"
+  input: "layer_512_1_conv2_h/convolution/SpaceToBatchND/block_shape"
+  input: "layer_512_1_conv2_h/convolution/SpaceToBatchND/paddings"
+}
+node {
+  name: "layer_512_1_conv2_h/convolution"
+  op: "Conv2D"
+  input: "layer_512_1_conv2_h/convolution/SpaceToBatchND"
+  input: "layer_512_1_conv2_h/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "VALID"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "layer_512_1_conv2_h/convolution/BatchToSpaceND"
+  op: "BatchToSpaceND"
+  input: "layer_512_1_conv2_h/convolution"
+  input: "layer_512_1_conv2_h/convolution/BatchToSpaceND/block_shape"
+  input: "layer_512_1_conv2_h/convolution/BatchToSpaceND/crops"
+}
+node {
+  name: "add_3"
+  op: "Add"
+  input: "layer_512_1_conv2_h/convolution/BatchToSpaceND"
+  input: "layer_512_1_conv_expand_h/Conv2D"
+}
+node {
+  name: "last_bn_h/FusedBatchNorm"
+  op: "FusedBatchNorm"
+  input: "add_3"
+  input: "last_bn_h/gamma"
+  input: "last_bn_h/beta"
+  input: "last_bn_h/mean"
+  input: "last_bn_h/std"
+  attr {
+    key: "epsilon"
+    value {
+      f: 1.00099996416e-05
+    }
+  }
+}
+node {
+  name: "last_scale_h/Mul"
+  op: "Mul"
+  input: "last_bn_h/FusedBatchNorm"
+  input: "last_scale_h/mul"
+}
+node {
+  name: "last_scale_h/BiasAdd"
+  op: "BiasAdd"
+  input: "last_scale_h/Mul"
+  input: "last_scale_h/add"
+}
+node {
+  name: "last_relu"
+  op: "Relu"
+  input: "last_scale_h/BiasAdd"
+}
+node {
+  name: "conv6_1_h/Conv2D"
+  op: "Conv2D"
+  input: "last_relu"
+  input: "conv6_1_h/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv6_1_h/BiasAdd"
+  op: "BiasAdd"
+  input: "conv6_1_h/Conv2D"
+  input: "conv6_1_h/bias"
+}
+node {
+  name: "conv6_1_h/Relu"
+  op: "Relu"
+  input: "conv6_1_h/BiasAdd"
+}
+node {
+  name: "conv6_2_h/Conv2D"
+  op: "Conv2D"
+  input: "conv6_1_h/Relu"
+  input: "conv6_2_h/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 2
+        i: 2
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv6_2_h/BiasAdd"
+  op: "BiasAdd"
+  input: "conv6_2_h/Conv2D"
+  input: "conv6_2_h/bias"
+}
+node {
+  name: "conv6_2_h/Relu"
+  op: "Relu"
+  input: "conv6_2_h/BiasAdd"
+}
+node {
+  name: "conv7_1_h/Conv2D"
+  op: "Conv2D"
+  input: "conv6_2_h/Relu"
+  input: "conv7_1_h/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv7_1_h/BiasAdd"
+  op: "BiasAdd"
+  input: "conv7_1_h/Conv2D"
+  input: "conv7_1_h/bias"
+}
+node {
+  name: "conv7_1_h/Relu"
+  op: "Relu"
+  input: "conv7_1_h/BiasAdd"
+}
+node {
+  name: "Pad_2"
+  op: "SpaceToBatchND"
+  input: "conv7_1_h/Relu"
+  input: "SpaceToBatchND/block_shape"
+  input: "SpaceToBatchND_1/paddings"
+}
+node {
+  name: "conv7_2_h/Conv2D"
+  op: "Conv2D"
+  input: "Pad_2"
+  input: "conv7_2_h/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "VALID"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 2
+        i: 2
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv7_2_h/BiasAdd"
+  op: "BiasAdd"
+  input: "conv7_2_h/Conv2D"
+  input: "conv7_2_h/bias"
+}
+node {
+  name: "BatchToSpaceND_2"
+  op: "BatchToSpaceND"
+  input: "conv7_2_h/BiasAdd"
+}
+node {
+  name: "conv7_2_h/Relu"
+  op: "Relu"
+  input: "BatchToSpaceND_2"
+}
+node {
+  name: "conv8_1_h/Conv2D"
+  op: "Conv2D"
+  input: "conv7_2_h/Relu"
+  input: "conv8_1_h/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv8_1_h/BiasAdd"
+  op: "BiasAdd"
+  input: "conv8_1_h/Conv2D"
+  input: "conv8_1_h/bias"
+}
+node {
+  name: "conv8_1_h/Relu"
+  op: "Relu"
+  input: "conv8_1_h/BiasAdd"
+}
+node {
+  name: "conv8_2_h/Conv2D"
+  op: "Conv2D"
+  input: "conv8_1_h/Relu"
+  input: "conv8_2_h/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv8_2_h/BiasAdd"
+  op: "BiasAdd"
+  input: "conv8_2_h/Conv2D"
+  input: "conv8_2_h/bias"
+}
+node {
+  name: "conv8_2_h/Relu"
+  op: "Relu"
+  input: "conv8_2_h/BiasAdd"
+}
+node {
+  name: "conv9_1_h/Conv2D"
+  op: "Conv2D"
+  input: "conv8_2_h/Relu"
+  input: "conv9_1_h/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv9_1_h/BiasAdd"
+  op: "BiasAdd"
+  input: "conv9_1_h/Conv2D"
+  input: "conv9_1_h/bias"
+}
+node {
+  name: "conv9_1_h/Relu"
+  op: "Relu"
+  input: "conv9_1_h/BiasAdd"
+}
+node {
+  name: "conv9_2_h/Conv2D"
+  op: "Conv2D"
+  input: "conv9_1_h/Relu"
+  input: "conv9_2_h/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv9_2_h/BiasAdd"
+  op: "BiasAdd"
+  input: "conv9_2_h/Conv2D"
+  input: "conv9_2_h/bias"
+}
+node {
+  name: "conv9_2_h/Relu"
+  op: "Relu"
+  input: "conv9_2_h/BiasAdd"
+}
+node {
+  name: "conv9_2_mbox_loc/Conv2D"
+  op: "Conv2D"
+  input: "conv9_2_h/Relu"
+  input: "conv9_2_mbox_loc/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv9_2_mbox_loc/BiasAdd"
+  op: "BiasAdd"
+  input: "conv9_2_mbox_loc/Conv2D"
+  input: "conv9_2_mbox_loc/bias"
+}
+node {
+  name: "flatten_5/Reshape"
+  op: "Flatten"
+  input: "conv9_2_mbox_loc/BiasAdd"
+}
+node {
+  name: "conv9_2_mbox_conf/Conv2D"
+  op: "Conv2D"
+  input: "conv9_2_h/Relu"
+  input: "conv9_2_mbox_conf/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv9_2_mbox_conf/BiasAdd"
+  op: "BiasAdd"
+  input: "conv9_2_mbox_conf/Conv2D"
+  input: "conv9_2_mbox_conf/bias"
+}
+node {
+  name: "flatten_11/Reshape"
+  op: "Flatten"
+  input: "conv9_2_mbox_conf/BiasAdd"
+}
+node {
+  name: "conv8_2_mbox_loc/Conv2D"
+  op: "Conv2D"
+  input: "conv8_2_h/Relu"
+  input: "conv8_2_mbox_loc/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv8_2_mbox_loc/BiasAdd"
+  op: "BiasAdd"
+  input: "conv8_2_mbox_loc/Conv2D"
+  input: "conv8_2_mbox_loc/bias"
+}
+node {
+  name: "flatten_4/Reshape"
+  op: "Flatten"
+  input: "conv8_2_mbox_loc/BiasAdd"
+}
+node {
+  name: "conv8_2_mbox_conf/Conv2D"
+  op: "Conv2D"
+  input: "conv8_2_h/Relu"
+  input: "conv8_2_mbox_conf/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv8_2_mbox_conf/BiasAdd"
+  op: "BiasAdd"
+  input: "conv8_2_mbox_conf/Conv2D"
+  input: "conv8_2_mbox_conf/bias"
+}
+node {
+  name: "flatten_10/Reshape"
+  op: "Flatten"
+  input: "conv8_2_mbox_conf/BiasAdd"
+}
+node {
+  name: "conv7_2_mbox_loc/Conv2D"
+  op: "Conv2D"
+  input: "conv7_2_h/Relu"
+  input: "conv7_2_mbox_loc/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv7_2_mbox_loc/BiasAdd"
+  op: "BiasAdd"
+  input: "conv7_2_mbox_loc/Conv2D"
+  input: "conv7_2_mbox_loc/bias"
+}
+node {
+  name: "flatten_3/Reshape"
+  op: "Flatten"
+  input: "conv7_2_mbox_loc/BiasAdd"
+}
+node {
+  name: "conv7_2_mbox_conf/Conv2D"
+  op: "Conv2D"
+  input: "conv7_2_h/Relu"
+  input: "conv7_2_mbox_conf/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv7_2_mbox_conf/BiasAdd"
+  op: "BiasAdd"
+  input: "conv7_2_mbox_conf/Conv2D"
+  input: "conv7_2_mbox_conf/bias"
+}
+node {
+  name: "flatten_9/Reshape"
+  op: "Flatten"
+  input: "conv7_2_mbox_conf/BiasAdd"
+}
+node {
+  name: "conv6_2_mbox_loc/Conv2D"
+  op: "Conv2D"
+  input: "conv6_2_h/Relu"
+  input: "conv6_2_mbox_loc/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv6_2_mbox_loc/BiasAdd"
+  op: "BiasAdd"
+  input: "conv6_2_mbox_loc/Conv2D"
+  input: "conv6_2_mbox_loc/bias"
+}
+node {
+  name: "flatten_2/Reshape"
+  op: "Flatten"
+  input: "conv6_2_mbox_loc/BiasAdd"
+}
+node {
+  name: "conv6_2_mbox_conf/Conv2D"
+  op: "Conv2D"
+  input: "conv6_2_h/Relu"
+  input: "conv6_2_mbox_conf/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "conv6_2_mbox_conf/BiasAdd"
+  op: "BiasAdd"
+  input: "conv6_2_mbox_conf/Conv2D"
+  input: "conv6_2_mbox_conf/bias"
+}
+node {
+  name: "flatten_8/Reshape"
+  op: "Flatten"
+  input: "conv6_2_mbox_conf/BiasAdd"
+}
+node {
+  name: "fc7_mbox_loc/Conv2D"
+  op: "Conv2D"
+  input: "last_relu"
+  input: "fc7_mbox_loc/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "fc7_mbox_loc/BiasAdd"
+  op: "BiasAdd"
+  input: "fc7_mbox_loc/Conv2D"
+  input: "fc7_mbox_loc/bias"
+}
+node {
+  name: "flatten_1/Reshape"
+  op: "Flatten"
+  input: "fc7_mbox_loc/BiasAdd"
+}
+node {
+  name: "mbox_loc"
+  op: "ConcatV2"
+  input: "flatten/Reshape"
+  input: "flatten_1/Reshape"
+  input: "flatten_2/Reshape"
+  input: "flatten_3/Reshape"
+  input: "flatten_4/Reshape"
+  input: "flatten_5/Reshape"
+  input: "mbox_loc/axis"
+}
+node {
+  name: "fc7_mbox_conf/Conv2D"
+  op: "Conv2D"
+  input: "last_relu"
+  input: "fc7_mbox_conf/weights"
+  attr {
+    key: "dilations"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+  attr {
+    key: "padding"
+    value {
+      s: "SAME"
+    }
+  }
+  attr {
+    key: "strides"
+    value {
+      list {
+        i: 1
+        i: 1
+        i: 1
+        i: 1
+      }
+    }
+  }
+}
+node {
+  name: "fc7_mbox_conf/BiasAdd"
+  op: "BiasAdd"
+  input: "fc7_mbox_conf/Conv2D"
+  input: "fc7_mbox_conf/bias"
+}
+node {
+  name: "flatten_7/Reshape"
+  op: "Flatten"
+  input: "fc7_mbox_conf/BiasAdd"
+}
+node {
+  name: "mbox_conf"
+  op: "ConcatV2"
+  input: "flatten_6/Reshape"
+  input: "flatten_7/Reshape"
+  input: "flatten_8/Reshape"
+  input: "flatten_9/Reshape"
+  input: "flatten_10/Reshape"
+  input: "flatten_11/Reshape"
+  input: "mbox_conf/axis"
+}
+node {
+  name: "mbox_conf_reshape"
+  op: "Reshape"
+  input: "mbox_conf"
+  input: "reshape_before_softmax"
+}
+node {
+  name: "mbox_conf_softmax"
+  op: "Softmax"
+  input: "mbox_conf_reshape"
+  attr {
+    key: "axis"
+    value {
+      i: 2
+    }
+  }
+}
+node {
+  name: "mbox_conf_flatten"
+  op: "Flatten"
+  input: "mbox_conf_softmax"
+}
+node {
+  name: "PriorBox_0"
+  op: "PriorBox"
+  input: "conv4_3_norm/mul_1"
+  input: "data"
+  attr {
+    key: "aspect_ratio"
+    value {
+      tensor {
+        dtype: DT_FLOAT
+        tensor_shape {
+          dim {
+            size: 1
+          }
+        }
+        float_val: 2.0
+      }
+    }
+  }
+  attr {
+    key: "clip"
+    value {
+      b: false
+    }
+  }
+  attr {
+    key: "flip"
+    value {
+      b: true
+    }
+  }
+  attr {
+    key: "max_size"
+    value {
+      i: 60
+    }
+  }
+  attr {
+    key: "min_size"
+    value {
+      i: 30
+    }
+  }
+  attr {
+    key: "offset"
+    value {
+      f: 0.5
+    }
+  }
+  attr {
+    key: "step"
+    value {
+      f: 8.0
+    }
+  }
+  attr {
+    key: "variance"
+    value {
+      tensor {
+        dtype: DT_FLOAT
+        tensor_shape {
+          dim {
+            size: 4
+          }
+        }
+        float_val: 0.10000000149
+        float_val: 0.10000000149
+        float_val: 0.20000000298
+        float_val: 0.20000000298
+      }
+    }
+  }
+}
+node {
+  name: "PriorBox_1"
+  op: "PriorBox"
+  input: "last_relu"
+  input: "data"
+  attr {
+    key: "aspect_ratio"
+    value {
+      tensor {
+        dtype: DT_FLOAT
+        tensor_shape {
+          dim {
+            size: 2
+          }
+        }
+        float_val: 2.0
+        float_val: 3.0
+      }
+    }
+  }
+  attr {
+    key: "clip"
+    value {
+      b: false
+    }
+  }
+  attr {
+    key: "flip"
+    value {
+      b: true
+    }
+  }
+  attr {
+    key: "max_size"
+    value {
+      i: 111
+    }
+  }
+  attr {
+    key: "min_size"
+    value {
+      i: 60
+    }
+  }
+  attr {
+    key: "offset"
+    value {
+      f: 0.5
+    }
+  }
+  attr {
+    key: "step"
+    value {
+      f: 16.0
+    }
+  }
+  attr {
+    key: "variance"
+    value {
+      tensor {
+        dtype: DT_FLOAT
+        tensor_shape {
+          dim {
+            size: 4
+          }
+        }
+        float_val: 0.10000000149
+        float_val: 0.10000000149
+        float_val: 0.20000000298
+        float_val: 0.20000000298
+      }
+    }
+  }
+}
+node {
+  name: "PriorBox_2"
+  op: "PriorBox"
+  input: "conv6_2_h/Relu"
+  input: "data"
+  attr {
+    key: "aspect_ratio"
+    value {
+      tensor {
+        dtype: DT_FLOAT
+        tensor_shape {
+          dim {
+            size: 2
+          }
+        }
+        float_val: 2.0
+        float_val: 3.0
+      }
+    }
+  }
+  attr {
+    key: "clip"
+    value {
+      b: false
+    }
+  }
+  attr {
+    key: "flip"
+    value {
+      b: true
+    }
+  }
+  attr {
+    key: "max_size"
+    value {
+      i: 162
+    }
+  }
+  attr {
+    key: "min_size"
+    value {
+      i: 111
+    }
+  }
+  attr {
+    key: "offset"
+    value {
+      f: 0.5
+    }
+  }
+  attr {
+    key: "step"
+    value {
+      f: 32.0
+    }
+  }
+  attr {
+    key: "variance"
+    value {
+      tensor {
+        dtype: DT_FLOAT
+        tensor_shape {
+          dim {
+            size: 4
+          }
+        }
+        float_val: 0.10000000149
+        float_val: 0.10000000149
+        float_val: 0.20000000298
+        float_val: 0.20000000298
+      }
+    }
+  }
+}
+node {
+  name: "PriorBox_3"
+  op: "PriorBox"
+  input: "conv7_2_h/Relu"
+  input: "data"
+  attr {
+    key: "aspect_ratio"
+    value {
+      tensor {
+        dtype: DT_FLOAT
+        tensor_shape {
+          dim {
+            size: 2
+          }
+        }
+        float_val: 2.0
+        float_val: 3.0
+      }
+    }
+  }
+  attr {
+    key: "clip"
+    value {
+      b: false
+    }
+  }
+  attr {
+    key: "flip"
+    value {
+      b: true
+    }
+  }
+  attr {
+    key: "max_size"
+    value {
+      i: 213
+    }
+  }
+  attr {
+    key: "min_size"
+    value {
+      i: 162
+    }
+  }
+  attr {
+    key: "offset"
+    value {
+      f: 0.5
+    }
+  }
+  attr {
+    key: "step"
+    value {
+      f: 64.0
+    }
+  }
+  attr {
+    key: "variance"
+    value {
+      tensor {
+        dtype: DT_FLOAT
+        tensor_shape {
+          dim {
+            size: 4
+          }
+        }
+        float_val: 0.10000000149
+        float_val: 0.10000000149
+        float_val: 0.20000000298
+        float_val: 0.20000000298
+      }
+    }
+  }
+}
+node {
+  name: "PriorBox_4"
+  op: "PriorBox"
+  input: "conv8_2_h/Relu"
+  input: "data"
+  attr {
+    key: "aspect_ratio"
+    value {
+      tensor {
+        dtype: DT_FLOAT
+        tensor_shape {
+          dim {
+            size: 1
+          }
+        }
+        float_val: 2.0
+      }
+    }
+  }
+  attr {
+    key: "clip"
+    value {
+      b: false
+    }
+  }
+  attr {
+    key: "flip"
+    value {
+      b: true
+    }
+  }
+  attr {
+    key: "max_size"
+    value {
+      i: 264
+    }
+  }
+  attr {
+    key: "min_size"
+    value {
+      i: 213
+    }
+  }
+  attr {
+    key: "offset"
+    value {
+      f: 0.5
+    }
+  }
+  attr {
+    key: "step"
+    value {
+      f: 100.0
+    }
+  }
+  attr {
+    key: "variance"
+    value {
+      tensor {
+        dtype: DT_FLOAT
+        tensor_shape {
+          dim {
+            size: 4
+          }
+        }
+        float_val: 0.10000000149
+        float_val: 0.10000000149
+        float_val: 0.20000000298
+        float_val: 0.20000000298
+      }
+    }
+  }
+}
+node {
+  name: "PriorBox_5"
+  op: "PriorBox"
+  input: "conv9_2_h/Relu"
+  input: "data"
+  attr {
+    key: "aspect_ratio"
+    value {
+      tensor {
+        dtype: DT_FLOAT
+        tensor_shape {
+          dim {
+            size: 1
+          }
+        }
+        float_val: 2.0
+      }
+    }
+  }
+  attr {
+    key: "clip"
+    value {
+      b: false
+    }
+  }
+  attr {
+    key: "flip"
+    value {
+      b: true
+    }
+  }
+  attr {
+    key: "max_size"
+    value {
+      i: 315
+    }
+  }
+  attr {
+    key: "min_size"
+    value {
+      i: 264
+    }
+  }
+  attr {
+    key: "offset"
+    value {
+      f: 0.5
+    }
+  }
+  attr {
+    key: "step"
+    value {
+      f: 300.0
+    }
+  }
+  attr {
+    key: "variance"
+    value {
+      tensor {
+        dtype: DT_FLOAT
+        tensor_shape {
+          dim {
+            size: 4
+          }
+        }
+        float_val: 0.10000000149
+        float_val: 0.10000000149
+        float_val: 0.20000000298
+        float_val: 0.20000000298
+      }
+    }
+  }
+}
+node {
+  name: "mbox_priorbox"
+  op: "ConcatV2"
+  input: "PriorBox_0"
+  input: "PriorBox_1"
+  input: "PriorBox_2"
+  input: "PriorBox_3"
+  input: "PriorBox_4"
+  input: "PriorBox_5"
+  input: "mbox_loc/axis"
+}
+node {
+  name: "detection_out"
+  op: "DetectionOutput"
+  input: "mbox_loc"
+  input: "mbox_conf_flatten"
+  input: "mbox_priorbox"
+  attr {
+    key: "background_label_id"
+    value {
+      i: 0
+    }
+  }
+  attr {
+    key: "code_type"
+    value {
+      s: "CENTER_SIZE"
+    }
+  }
+  attr {
+    key: "confidence_threshold"
+    value {
+      f: 0.00999999977648
+    }
+  }
+  attr {
+    key: "keep_top_k"
+    value {
+      i: 200
+    }
+  }
+  attr {
+    key: "nms_threshold"
+    value {
+      f: 0.449999988079
+    }
+  }
+  attr {
+    key: "num_classes"
+    value {
+      i: 2
+    }
+  }
+  attr {
+    key: "share_location"
+    value {
+      b: true
+    }
+  }
+  attr {
+    key: "top_k"
+    value {
+      i: 400
+    }
+  }
+}
+node {
+  name: "reshape_before_softmax"
+  op: "Const"
+  attr {
+    key: "value"
+    value {
+      tensor {
+        dtype: DT_INT32
+        tensor_shape {
+          dim {
+            size: 3
+          }
+        }
+        int_val: 0
+        int_val: -1
+        int_val: 2
+      }
+    }
+  }
+}
+library {
+}

opencv_face_detector_uint8.pb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c71d752ef2cbf2f457ac82fdd580fcb2522fd04c5efdaed18eb6d9e2843fbed
+size 2727750