mirror of
https://github.com/opencv/opencv.git
synced 2026-07-29 23:33:05 +04:00
Merge remote-tracking branch 'upstream/3.4' into merge-3.4
Revert "documentation: avoid links to 'master' branch from 3.4 maintenance branch" This reverts commit9ba9358ecb. Revert "documentation: avoid links to 'master' branch from 3.4 maintenance branch (2)" This reverts commitf185802489.
This commit is contained in:
Alexander Alekhin
@@ -11,8 +11,10 @@
|
||||
| [SSDs from TensorFlow](https://github.com/tensorflow/models/tree/master/research/object_detection/) | `0.00784 (2/255)` | `300x300` | `127.5 127.5 127.5` | RGB |
|
||||
| [YOLO](https://pjreddie.com/darknet/yolo/) | `0.00392 (1/255)` | `416x416` | `0 0 0` | RGB |
|
||||
| [VGG16-SSD](https://github.com/weiliu89/caffe/tree/ssd) | `1.0` | `300x300` | `104 117 123` | BGR |
|
||||
| [Faster-RCNN](https://github.com/rbgirshick/py-faster-rcnn) | `1.0` | `800x600` | `102.9801, 115.9465, 122.7717` | BGR |
|
||||
| [Faster-RCNN](https://github.com/rbgirshick/py-faster-rcnn) | `1.0` | `800x600` | `102.9801 115.9465 122.7717` | BGR |
|
||||
| [R-FCN](https://github.com/YuwenXiong/py-R-FCN) | `1.0` | `800x600` | `102.9801 115.9465 122.7717` | BGR |
|
||||
| [Faster-RCNN, ResNet backbone](https://github.com/tensorflow/models/tree/master/research/object_detection/) | `1.0` | `300x300` | `103.939 116.779 123.68` | RGB |
|
||||
| [Faster-RCNN, InceptionV2 backbone](https://github.com/tensorflow/models/tree/master/research/object_detection/) | `0.00784 (2/255)` | `300x300` | `127.5 127.5 127.5` | RGB |
|
||||
|
||||
#### Face detection
|
||||
[An origin model](https://github.com/opencv/opencv/tree/master/samples/dnn/face_detector)
|
||||
|
||||
@@ -23,7 +23,7 @@ const char* keys =
|
||||
"{ backend | 0 | Choose one of computation backends: "
|
||||
"0: default C++ backend, "
|
||||
"1: Halide language (http://halide-lang.org/), "
|
||||
"2: Intel's Deep Learning Inference Engine (https://software.seek.intel.com/deep-learning-deployment)}"
|
||||
"2: Intel's Deep Learning Inference Engine (https://software.intel.com/openvino-toolkit)}"
|
||||
"{ target | 0 | Choose one of target computation devices: "
|
||||
"0: CPU target (by default),"
|
||||
"1: OpenCL }";
|
||||
|
||||
@@ -34,7 +34,7 @@ parser.add_argument('--backend', choices=backends, default=cv.dnn.DNN_BACKEND_DE
|
||||
help="Choose one of computation backends: "
|
||||
"%d: default C++ backend, "
|
||||
"%d: Halide language (http://halide-lang.org/), "
|
||||
"%d: Intel's Deep Learning Inference Engine (https://software.seek.intel.com/deep-learning-deployment)" % backends)
|
||||
"%d: Intel's Deep Learning Inference Engine (https://software.intel.com/openvino-toolkit)" % backends)
|
||||
parser.add_argument('--target', choices=targets, default=cv.dnn.DNN_TARGET_CPU, type=int,
|
||||
help='Choose one of target computation devices: '
|
||||
'%d: CPU target (by default), '
|
||||
|
||||
@@ -13,7 +13,7 @@ The data preparation pipeline can be represented as:
|
||||
a) Find some datasets with face bounding boxes annotation. For some reasons I can't provide links here, but you easily find them on your own. Also study the data. It may contain small or low quality faces which can spoil training process. Often there are special flags about object quality in annotation. Remove such faces from annotation (smaller when 16 along at least one side, or blurred, of highly-occluded, or something else).
|
||||
|
||||
b) The downloaded dataset will have some format of annotation. It may be one single file for all images, or separate file for each image or something else. But to train SSD in Caffe you need to convert annotation to PASCAL VOC format.
|
||||
PASCAL VOC annoitation consist of .xml file for each image. In this xml file all face bounding boxes should be listed as:
|
||||
PASCAL VOC annotation consist of .xml file for each image. In this xml file all face bounding boxes should be listed as:
|
||||
|
||||
<annotation>
|
||||
<size>
|
||||
@@ -42,7 +42,7 @@ PASCAL VOC annoitation consist of .xml file for each image. In this xml file all
|
||||
</object>
|
||||
</annotation>
|
||||
|
||||
So, convert your dataset's annotation to the fourmat above.
|
||||
So, convert your dataset's annotation to the format above.
|
||||
Also, you should create labelmap.prototxt file with the following content:
|
||||
item {
|
||||
name: "none_of_the_above"
|
||||
@@ -76,4 +76,4 @@ mkdir -p log
|
||||
/path_for_caffe_build_dir/tools/caffe train -solver="solver.prototxt" -gpu 0 2>&1 | tee -a log/log.log
|
||||
|
||||
And wait. It will take about 8 hours to finish the process.
|
||||
After it you can use your .caffemodel from snapshot/ subdirectory in resnet_face_ssd_python.py sample.
|
||||
After it you can use your .caffemodel from snapshot/ subdirectory in resnet_face_ssd_python.py sample.
|
||||
|
||||
@@ -25,7 +25,7 @@ const char* keys =
|
||||
"{ backend | 0 | Choose one of computation backends: "
|
||||
"0: default C++ backend, "
|
||||
"1: Halide language (http://halide-lang.org/), "
|
||||
"2: Intel's Deep Learning Inference Engine (https://software.seek.intel.com/deep-learning-deployment)}"
|
||||
"2: Intel's Deep Learning Inference Engine (https://software.intel.com/openvino-toolkit)}"
|
||||
"{ target | 0 | Choose one of target computation devices: "
|
||||
"0: CPU target (by default),"
|
||||
"1: OpenCL }";
|
||||
|
||||
@@ -35,7 +35,7 @@ parser.add_argument('--backend', choices=backends, default=cv.dnn.DNN_BACKEND_DE
|
||||
help="Choose one of computation backends: "
|
||||
"%d: default C++ backend, "
|
||||
"%d: Halide language (http://halide-lang.org/), "
|
||||
"%d: Intel's Deep Learning Inference Engine (https://software.seek.intel.com/deep-learning-deployment)" % backends)
|
||||
"%d: Intel's Deep Learning Inference Engine (https://software.intel.com/openvino-toolkit)" % backends)
|
||||
parser.add_argument('--target', choices=targets, default=cv.dnn.DNN_TARGET_CPU, type=int,
|
||||
help='Choose one of target computation devices: '
|
||||
'%d: CPU target (by default), '
|
||||
@@ -174,7 +174,7 @@ while cv.waitKey(1) < 0:
|
||||
net.setInput(blob)
|
||||
if net.getLayer(0).outputNameToIndex('im_info') != -1: # Faster-RCNN or R-FCN
|
||||
frame = cv.resize(frame, (inpWidth, inpHeight))
|
||||
net.setInput(np.array([inpHeight, inpWidth, 1.6], dtype=np.float32), 'im_info');
|
||||
net.setInput(np.array([inpHeight, inpWidth, 1.6], dtype=np.float32), 'im_info')
|
||||
outs = net.forward(getOutputsNames(net))
|
||||
|
||||
postprocess(frame, outs)
|
||||
|
||||
@@ -26,7 +26,7 @@ const char* keys =
|
||||
"{ backend | 0 | Choose one of computation backends: "
|
||||
"0: default C++ backend, "
|
||||
"1: Halide language (http://halide-lang.org/), "
|
||||
"2: Intel's Deep Learning Inference Engine (https://software.seek.intel.com/deep-learning-deployment)}"
|
||||
"2: Intel's Deep Learning Inference Engine (https://software.intel.com/openvino-toolkit)}"
|
||||
"{ target | 0 | Choose one of target computation devices: "
|
||||
"0: CPU target (by default),"
|
||||
"1: OpenCL }";
|
||||
|
||||
@@ -36,7 +36,7 @@ parser.add_argument('--backend', choices=backends, default=cv.dnn.DNN_BACKEND_DE
|
||||
help="Choose one of computation backends: "
|
||||
"%d: default C++ backend, "
|
||||
"%d: Halide language (http://halide-lang.org/), "
|
||||
"%d: Intel's Deep Learning Inference Engine (https://software.seek.intel.com/deep-learning-deployment)" % backends)
|
||||
"%d: Intel's Deep Learning Inference Engine (https://software.intel.com/openvino-toolkit)" % backends)
|
||||
parser.add_argument('--target', choices=targets, default=cv.dnn.DNN_TARGET_CPU, type=int,
|
||||
help='Choose one of target computation devices: '
|
||||
'%d: CPU target (by default), '
|
||||
|
||||
@@ -0,0 +1,291 @@
|
||||
import argparse
|
||||
import numpy as np
|
||||
import tensorflow as tf
|
||||
|
||||
from tensorflow.core.framework.node_def_pb2 import NodeDef
|
||||
from tensorflow.tools.graph_transforms import TransformGraph
|
||||
from google.protobuf import text_format
|
||||
|
||||
parser = argparse.ArgumentParser(description='Run this script to get a text graph of '
|
||||
'SSD model from TensorFlow Object Detection API. '
|
||||
'Then pass it with .pb file to cv::dnn::readNetFromTensorflow function.')
|
||||
parser.add_argument('--input', required=True, help='Path to frozen TensorFlow graph.')
|
||||
parser.add_argument('--output', required=True, help='Path to output text graph.')
|
||||
parser.add_argument('--num_classes', default=90, type=int, help='Number of trained classes.')
|
||||
parser.add_argument('--scales', default=[0.25, 0.5, 1.0, 2.0], type=float, nargs='+',
|
||||
help='Hyper-parameter of grid_anchor_generator from a config file.')
|
||||
parser.add_argument('--aspect_ratios', default=[0.5, 1.0, 2.0], type=float, nargs='+',
|
||||
help='Hyper-parameter of grid_anchor_generator from a config file.')
|
||||
parser.add_argument('--features_stride', default=16, type=float, nargs='+',
|
||||
help='Hyper-parameter from a config file.')
|
||||
args = parser.parse_args()
|
||||
|
||||
scopesToKeep = ('FirstStageFeatureExtractor', 'Conv',
|
||||
'FirstStageBoxPredictor/BoxEncodingPredictor',
|
||||
'FirstStageBoxPredictor/ClassPredictor',
|
||||
'CropAndResize',
|
||||
'MaxPool2D',
|
||||
'SecondStageFeatureExtractor',
|
||||
'SecondStageBoxPredictor',
|
||||
'image_tensor')
|
||||
|
||||
scopesToIgnore = ('FirstStageFeatureExtractor/Assert',
|
||||
'FirstStageFeatureExtractor/Shape',
|
||||
'FirstStageFeatureExtractor/strided_slice',
|
||||
'FirstStageFeatureExtractor/GreaterEqual',
|
||||
'FirstStageFeatureExtractor/LogicalAnd')
|
||||
|
||||
unusedAttrs = ['T', 'Tshape', 'N', 'Tidx', 'Tdim', 'use_cudnn_on_gpu',
|
||||
'Index', 'Tperm', 'is_training', 'Tpaddings']
|
||||
|
||||
# Read the graph.
|
||||
with tf.gfile.FastGFile(args.input, 'rb') as f:
|
||||
graph_def = tf.GraphDef()
|
||||
graph_def.ParseFromString(f.read())
|
||||
|
||||
# Removes Identity nodes
|
||||
def removeIdentity():
|
||||
identities = {}
|
||||
for node in graph_def.node:
|
||||
if node.op == 'Identity':
|
||||
identities[node.name] = node.input[0]
|
||||
graph_def.node.remove(node)
|
||||
|
||||
for node in graph_def.node:
|
||||
for i in range(len(node.input)):
|
||||
if node.input[i] in identities:
|
||||
node.input[i] = identities[node.input[i]]
|
||||
|
||||
removeIdentity()
|
||||
|
||||
removedNodes = []
|
||||
|
||||
for i in reversed(range(len(graph_def.node))):
|
||||
op = graph_def.node[i].op
|
||||
name = graph_def.node[i].name
|
||||
|
||||
if op == 'Const' or name.startswith(scopesToIgnore) or not name.startswith(scopesToKeep):
|
||||
if op != 'Const':
|
||||
removedNodes.append(name)
|
||||
|
||||
del graph_def.node[i]
|
||||
else:
|
||||
for attr in unusedAttrs:
|
||||
if attr in graph_def.node[i].attr:
|
||||
del graph_def.node[i].attr[attr]
|
||||
|
||||
# Remove references to removed nodes except Const nodes.
|
||||
for node in graph_def.node:
|
||||
for i in reversed(range(len(node.input))):
|
||||
if node.input[i] in removedNodes:
|
||||
del node.input[i]
|
||||
|
||||
|
||||
# Connect input node to the first layer
|
||||
assert(graph_def.node[0].op == 'Placeholder')
|
||||
graph_def.node[1].input.insert(0, graph_def.node[0].name)
|
||||
|
||||
# Temporarily remove top nodes.
|
||||
topNodes = []
|
||||
while True:
|
||||
node = graph_def.node.pop()
|
||||
topNodes.append(node)
|
||||
if node.op == 'CropAndResize':
|
||||
break
|
||||
|
||||
def tensorMsg(values):
|
||||
if all([isinstance(v, float) for v in values]):
|
||||
dtype = 'DT_FLOAT'
|
||||
field = 'float_val'
|
||||
elif all([isinstance(v, int) for v in values]):
|
||||
dtype = 'DT_INT32'
|
||||
field = 'int_val'
|
||||
else:
|
||||
raise Exception('Wrong values types')
|
||||
|
||||
msg = 'tensor { dtype: ' + dtype + ' tensor_shape { dim { size: %d } }' % len(values)
|
||||
for value in values:
|
||||
msg += '%s: %s ' % (field, str(value))
|
||||
return msg + '}'
|
||||
|
||||
def addSlice(inp, out, begins, sizes):
|
||||
beginsNode = NodeDef()
|
||||
beginsNode.name = out + '/begins'
|
||||
beginsNode.op = 'Const'
|
||||
text_format.Merge(tensorMsg(begins), beginsNode.attr["value"])
|
||||
graph_def.node.extend([beginsNode])
|
||||
|
||||
sizesNode = NodeDef()
|
||||
sizesNode.name = out + '/sizes'
|
||||
sizesNode.op = 'Const'
|
||||
text_format.Merge(tensorMsg(sizes), sizesNode.attr["value"])
|
||||
graph_def.node.extend([sizesNode])
|
||||
|
||||
sliced = NodeDef()
|
||||
sliced.name = out
|
||||
sliced.op = 'Slice'
|
||||
sliced.input.append(inp)
|
||||
sliced.input.append(beginsNode.name)
|
||||
sliced.input.append(sizesNode.name)
|
||||
graph_def.node.extend([sliced])
|
||||
|
||||
def addReshape(inp, out, shape):
|
||||
shapeNode = NodeDef()
|
||||
shapeNode.name = out + '/shape'
|
||||
shapeNode.op = 'Const'
|
||||
text_format.Merge(tensorMsg(shape), shapeNode.attr["value"])
|
||||
graph_def.node.extend([shapeNode])
|
||||
|
||||
reshape = NodeDef()
|
||||
reshape.name = out
|
||||
reshape.op = 'Reshape'
|
||||
reshape.input.append(inp)
|
||||
reshape.input.append(shapeNode.name)
|
||||
graph_def.node.extend([reshape])
|
||||
|
||||
def addSoftMax(inp, out):
|
||||
softmax = NodeDef()
|
||||
softmax.name = out
|
||||
softmax.op = 'Softmax'
|
||||
text_format.Merge('i: -1', softmax.attr['axis'])
|
||||
softmax.input.append(inp)
|
||||
graph_def.node.extend([softmax])
|
||||
|
||||
addReshape('FirstStageBoxPredictor/ClassPredictor/BiasAdd',
|
||||
'FirstStageBoxPredictor/ClassPredictor/reshape_1', [0, -1, 2])
|
||||
|
||||
addSoftMax('FirstStageBoxPredictor/ClassPredictor/reshape_1',
|
||||
'FirstStageBoxPredictor/ClassPredictor/softmax') # Compare with Reshape_4
|
||||
|
||||
flatten = NodeDef()
|
||||
flatten.name = 'FirstStageBoxPredictor/BoxEncodingPredictor/flatten' # Compare with FirstStageBoxPredictor/BoxEncodingPredictor/BiasAdd
|
||||
flatten.op = 'Flatten'
|
||||
flatten.input.append('FirstStageBoxPredictor/BoxEncodingPredictor/BiasAdd')
|
||||
graph_def.node.extend([flatten])
|
||||
|
||||
proposals = NodeDef()
|
||||
proposals.name = 'proposals' # Compare with ClipToWindow/Gather/Gather (NOTE: normalized)
|
||||
proposals.op = 'PriorBox'
|
||||
proposals.input.append('FirstStageBoxPredictor/BoxEncodingPredictor/BiasAdd')
|
||||
proposals.input.append(graph_def.node[0].name) # image_tensor
|
||||
|
||||
text_format.Merge('b: false', proposals.attr["flip"])
|
||||
text_format.Merge('b: true', proposals.attr["clip"])
|
||||
text_format.Merge('f: %f' % args.features_stride, proposals.attr["step"])
|
||||
text_format.Merge('f: 0.0', proposals.attr["offset"])
|
||||
text_format.Merge(tensorMsg([0.1, 0.1, 0.2, 0.2]), proposals.attr["variance"])
|
||||
|
||||
widths = []
|
||||
heights = []
|
||||
for a in args.aspect_ratios:
|
||||
for s in args.scales:
|
||||
ar = np.sqrt(a)
|
||||
heights.append((args.features_stride**2) * s / ar)
|
||||
widths.append((args.features_stride**2) * s * ar)
|
||||
|
||||
text_format.Merge(tensorMsg(widths), proposals.attr["width"])
|
||||
text_format.Merge(tensorMsg(heights), proposals.attr["height"])
|
||||
|
||||
graph_def.node.extend([proposals])
|
||||
|
||||
# Compare with Reshape_5
|
||||
detectionOut = NodeDef()
|
||||
detectionOut.name = 'detection_out'
|
||||
detectionOut.op = 'DetectionOutput'
|
||||
|
||||
detectionOut.input.append('FirstStageBoxPredictor/BoxEncodingPredictor/flatten')
|
||||
detectionOut.input.append('FirstStageBoxPredictor/ClassPredictor/softmax')
|
||||
detectionOut.input.append('proposals')
|
||||
|
||||
text_format.Merge('i: 2', detectionOut.attr['num_classes'])
|
||||
text_format.Merge('b: true', detectionOut.attr['share_location'])
|
||||
text_format.Merge('i: 0', detectionOut.attr['background_label_id'])
|
||||
text_format.Merge('f: 0.7', detectionOut.attr['nms_threshold'])
|
||||
text_format.Merge('i: 6000', detectionOut.attr['top_k'])
|
||||
text_format.Merge('s: "CENTER_SIZE"', detectionOut.attr['code_type'])
|
||||
text_format.Merge('i: 100', detectionOut.attr['keep_top_k'])
|
||||
text_format.Merge('b: true', detectionOut.attr['clip'])
|
||||
text_format.Merge('b: true', detectionOut.attr['loc_pred_transposed'])
|
||||
|
||||
graph_def.node.extend([detectionOut])
|
||||
|
||||
# Save as text.
|
||||
for node in reversed(topNodes):
|
||||
graph_def.node.extend([node])
|
||||
|
||||
addSoftMax('SecondStageBoxPredictor/Reshape_1', 'SecondStageBoxPredictor/Reshape_1/softmax')
|
||||
|
||||
addSlice('SecondStageBoxPredictor/Reshape_1/softmax',
|
||||
'SecondStageBoxPredictor/Reshape_1/slice',
|
||||
[0, 0, 1], [-1, -1, -1])
|
||||
|
||||
addReshape('SecondStageBoxPredictor/Reshape_1/slice',
|
||||
'SecondStageBoxPredictor/Reshape_1/Reshape', [1, -1])
|
||||
|
||||
# Replace Flatten subgraph onto a single node.
|
||||
for i in reversed(range(len(graph_def.node))):
|
||||
if graph_def.node[i].op == 'CropAndResize':
|
||||
graph_def.node[i].input.insert(1, 'detection_out')
|
||||
|
||||
if graph_def.node[i].name == 'SecondStageBoxPredictor/Reshape':
|
||||
shapeNode = NodeDef()
|
||||
shapeNode.name = 'SecondStageBoxPredictor/Reshape/shape2'
|
||||
shapeNode.op = 'Const'
|
||||
text_format.Merge(tensorMsg([1, -1, 4]), shapeNode.attr["value"])
|
||||
graph_def.node.extend([shapeNode])
|
||||
|
||||
graph_def.node[i].input.pop()
|
||||
graph_def.node[i].input.append(shapeNode.name)
|
||||
|
||||
if graph_def.node[i].name in ['SecondStageBoxPredictor/Flatten/flatten/Shape',
|
||||
'SecondStageBoxPredictor/Flatten/flatten/strided_slice',
|
||||
'SecondStageBoxPredictor/Flatten/flatten/Reshape/shape']:
|
||||
del graph_def.node[i]
|
||||
|
||||
for node in graph_def.node:
|
||||
if node.name == 'SecondStageBoxPredictor/Flatten/flatten/Reshape':
|
||||
node.op = 'Flatten'
|
||||
node.input.pop()
|
||||
break
|
||||
|
||||
################################################################################
|
||||
### Postprocessing
|
||||
################################################################################
|
||||
addSlice('detection_out', 'detection_out/slice', [0, 0, 0, 3], [-1, -1, -1, 4])
|
||||
|
||||
variance = NodeDef()
|
||||
variance.name = 'proposals/variance'
|
||||
variance.op = 'Const'
|
||||
text_format.Merge(tensorMsg([0.1, 0.1, 0.2, 0.2]), variance.attr["value"])
|
||||
graph_def.node.extend([variance])
|
||||
|
||||
varianceEncoder = NodeDef()
|
||||
varianceEncoder.name = 'variance_encoded'
|
||||
varianceEncoder.op = 'Mul'
|
||||
varianceEncoder.input.append('SecondStageBoxPredictor/Reshape')
|
||||
varianceEncoder.input.append(variance.name)
|
||||
text_format.Merge('i: 2', varianceEncoder.attr["axis"])
|
||||
graph_def.node.extend([varianceEncoder])
|
||||
|
||||
addReshape('detection_out/slice', 'detection_out/slice/reshape', [1, 1, -1])
|
||||
|
||||
detectionOut = NodeDef()
|
||||
detectionOut.name = 'detection_out_final'
|
||||
detectionOut.op = 'DetectionOutput'
|
||||
|
||||
detectionOut.input.append('variance_encoded')
|
||||
detectionOut.input.append('SecondStageBoxPredictor/Reshape_1/Reshape')
|
||||
detectionOut.input.append('detection_out/slice/reshape')
|
||||
|
||||
text_format.Merge('i: %d' % args.num_classes, detectionOut.attr['num_classes'])
|
||||
text_format.Merge('b: false', detectionOut.attr['share_location'])
|
||||
text_format.Merge('i: %d' % (args.num_classes + 1), detectionOut.attr['background_label_id'])
|
||||
text_format.Merge('f: 0.6', detectionOut.attr['nms_threshold'])
|
||||
text_format.Merge('s: "CENTER_SIZE"', detectionOut.attr['code_type'])
|
||||
text_format.Merge('i: 100', detectionOut.attr['keep_top_k'])
|
||||
text_format.Merge('b: true', detectionOut.attr['loc_pred_transposed'])
|
||||
text_format.Merge('b: true', detectionOut.attr['clip'])
|
||||
text_format.Merge('b: true', detectionOut.attr['variance_encoded_in_target'])
|
||||
graph_def.node.extend([detectionOut])
|
||||
|
||||
tf.train.write_graph(graph_def, "", args.output, as_text=True)
|
||||
Reference in New Issue
Block a user