History

Dmitry Kurtaev f8398d80bc add Net::getUnconnectedOutLayersNames method		2018-09-25 18:10:45 +03:00
..
face_detector	Update face detection network in samples	2018-08-14 13:16:23 +03:00
classification.cpp	core: repair CV_Assert() messages	2018-08-15 17:43:10 +03:00
classification.py	Make Intel's Inference Engine backend is default if no preferable backend is specified.	2018-06-04 18:31:46 +03:00
CMakeLists.txt	Install data for samples to correct directories, do not download face_detector models in cmake	2018-08-02 14:24:05 +03:00
colorization.cpp	Make Intel's Inference Engine backend is default if no preferable backend is specified.	2018-06-04 18:31:46 +03:00
colorization.py	Make Intel's Inference Engine backend is default if no preferable backend is specified.	2018-06-04 18:31:46 +03:00
custom_layers.hpp	Merge pull request #12264 from dkurt:dnn_remove_forward_method	2018-09-06 13:26:47 +03:00
edge_detection.py	Custom deep learning layers in Python	2018-04-26 09:25:18 +03:00
fast_neural_style.py	Make Intel's Inference Engine backend is default if no preferable backend is specified.	2018-06-04 18:31:46 +03:00
js_face_recognition.html	documentation: avoid links to 'master' branch from 3.4 maintenance branch (2)	2018-05-31 19:30:56 +03:00
mask_rcnn.py	Merge pull request #12243 from dkurt:dnn_tf_mask_rcnn	2018-08-24 14:47:32 +03:00
mobilenet_ssd_accuracy.py	Make Intel's Inference Engine backend is default if no preferable backend is specified.	2018-06-04 18:31:46 +03:00
object_detection.cpp	add Net::getUnconnectedOutLayersNames method	2018-09-25 18:10:45 +03:00
object_detection.py	add Net::getUnconnectedOutLayersNames method	2018-09-25 18:10:45 +03:00
openpose.cpp	select the device (video capture)	2018-05-09 17:20:02 +03:00
openpose.py	Make Intel's Inference Engine backend is default if no preferable backend is specified.	2018-06-04 18:31:46 +03:00
README.md	documentation: avoid links to 'master' branch from 3.4 maintenance branch (2)	2018-05-31 19:30:56 +03:00
segmentation.cpp	core: repair CV_Assert() messages	2018-08-15 17:43:10 +03:00
segmentation.py	Make Intel's Inference Engine backend is default if no preferable backend is specified.	2018-06-04 18:31:46 +03:00
shrink_tf_graph_weights.py	Text TensorFlow graphs parsing. MobileNet-SSD for 90 classes.	2017-10-08 22:25:29 +03:00
text_detection.cpp	core: repair CV_Assert() messages	2018-08-15 17:43:10 +03:00
tf_text_graph_common.py	Import tensorflow to create text graphs if import cv is failed	2018-09-18 09:11:51 +03:00
tf_text_graph_faster_rcnn.py	Import tensorflow to create text graphs if import cv is failed	2018-09-18 09:11:51 +03:00
tf_text_graph_mask_rcnn.py	Import tensorflow to create text graphs if import cv is failed	2018-09-18 09:11:51 +03:00
tf_text_graph_ssd.py	Import tensorflow to create text graphs if import cv is failed	2018-09-18 09:11:51 +03:00

README.md

OpenCV deep learning module samples

Model Zoo

Object detection

Model	Scale	Size WxH	Mean subtraction	Channels order
MobileNet-SSD, Caffe	`0.00784 (2/255)`	`300x300`	`127.5 127.5 127.5`	BGR
OpenCV face detector	`1.0`	`300x300`	`104 177 123`	BGR
SSDs from TensorFlow	`0.00784 (2/255)`	`300x300`	`127.5 127.5 127.5`	RGB
YOLO	`0.00392 (1/255)`	`416x416`	`0 0 0`	RGB
VGG16-SSD	`1.0`	`300x300`	`104 117 123`	BGR
Faster-RCNN	`1.0`	`800x600`	`102.9801 115.9465 122.7717`	BGR
R-FCN	`1.0`	`800x600`	`102.9801 115.9465 122.7717`	BGR
Faster-RCNN, ResNet backbone	`1.0`	`300x300`	`103.939 116.779 123.68`	RGB
Faster-RCNN, InceptionV2 backbone	`0.00784 (2/255)`	`300x300`	`127.5 127.5 127.5`	RGB

Face detection

An origin model with single precision floating point weights has been quantized using TensorFlow framework. To achieve the best accuracy run the model on BGR images resized to 300x300 applying mean subtraction of values (104, 177, 123) for each blue, green and red channels correspondingly.

The following are accuracy metrics obtained using COCO object detection evaluation tool on FDDB dataset (see script) applying resize to 300x300 and keeping an origin images' sizes.

AP - Average Precision                            | FP32/FP16 | UINT8          | FP32/FP16 | UINT8          |
AR - Average Recall                               | 300x300   | 300x300        | any size  | any size       |
--------------------------------------------------|-----------|----------------|-----------|----------------|
AP @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] | 0.408     | 0.408          | 0.378     | 0.328 (-0.050) |
AP @[ IoU=0.50      | area=   all | maxDets=100 ] | 0.849     | 0.849          | 0.797     | 0.790 (-0.007) |
AP @[ IoU=0.75      | area=   all | maxDets=100 ] | 0.251     | 0.251          | 0.208     | 0.140 (-0.068) |
AP @[ IoU=0.50:0.95 | area= small | maxDets=100 ] | 0.050     | 0.051 (+0.001) | 0.107     | 0.070 (-0.037) |
AP @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] | 0.381     | 0.379 (-0.002) | 0.380     | 0.368 (-0.012) |
AP @[ IoU=0.50:0.95 | area= large | maxDets=100 ] | 0.455     | 0.455          | 0.412     | 0.337 (-0.075) |
AR @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] | 0.299     | 0.299          | 0.279     | 0.246 (-0.033) |
AR @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] | 0.482     | 0.482          | 0.476     | 0.436 (-0.040) |
AR @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] | 0.496     | 0.496          | 0.491     | 0.451 (-0.040) |
AR @[ IoU=0.50:0.95 | area= small | maxDets=100 ] | 0.189     | 0.193 (+0.004) | 0.284     | 0.232 (-0.052) |
AR @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] | 0.481     | 0.480 (-0.001) | 0.470     | 0.458 (-0.012) |
AR @[ IoU=0.50:0.95 | area= large | maxDets=100 ] | 0.528     | 0.528          | 0.520     | 0.462 (-0.058) |

Classification

Model	Scale	Size WxH	Mean subtraction	Channels order
GoogLeNet	`1.0`	`224x224`	`104 117 123`	BGR
SqueezeNet	`1.0`	`227x227`	`0 0 0`	BGR

Semantic segmentation

Model	Scale	Size WxH	Mean subtraction	Channels order
ENet	`0.00392 (1/255)`	`1024x512`	`0 0 0`	RGB
FCN8s	`1.0`	`500x500`	`0 0 0`	BGR