OpenCV3.4.2で始める浅いディープラーニング (その6 その他のexample)

(編集中)
順番がいきなり(その6)になってますが、気にしないでください。
その他のexampleの実行方法を確認していきます。

４．example_dnn_openpose.exe

人体の骨格認識をしてくれます。入力は2次元画像で、出力も2次元です。

A sample app to demonstrate human or hand pose detection with a pretrained OpenPose dnn.
Usage: example_dnn_openpose.exe [params]
-h, --help (value:false)
print this help message
--height (value:368)
Preprocess input image by resizing to a specific height.
-i, --image
(required) path to image file (containing a single person, or hand)
-m, --model
(required) model weights, e.g. hand/pose_iter_102000.caffemodel
-p, --proto
(required) model configuration, e.g. hand/pose.prototxt
-t, --threshold (value:0.1)
threshold or confidence value for the heatmap
--width (value:368)
Preprocess input image by resizing to a specific width.

実行例

example_dnn_openpose.exe --model=openpose_pose_coco.caffemodel --proto=openpose_pose_coco.prototxt --image=googlenet_0.png

f:id:cvl-robot:20180719211646p:plain

５．example_dnn_colorization.exe

グレースケールの画像の自動色付け。

This sample demonstrates recoloring grayscale images with dnn.
This program is based on:
http://richzhang.github.io/colorization
https://github.com/richzhang/colorization
Download caffemodel and prototxt files:
http://eecs.berkeley.edu/~rich.zhang/projects/2016_colorization/files/demo_v2/colorization_release_v2.caffemodel
https://raw.githubusercontent.com/richzhang/colorization/master/colorization/models/colorization_deploy_v2.prototxt
Usage: example_dnn_colorization.exe [params]
-h, --help (value:true)
print this help message
--image (value:space_shuttle.jpg)
path to image file
--model (value:colorization_release_v2.caffemodel)
model weights
--opencl
enable OpenCL
--proto (value:colorization_deploy_v2.prototxt)
model configuration

実行例

example_dnn_colorization.exe

f:id:cvl-robot:20180719210619p:plain

６．example_dnn_text_detection.exe

文字検出。日本語も。

Use this script to run TensorFlow implementation (https://github.com/argman/EAST) of EAST: An Efficient and Accurate Scene Text Detector (https://arxiv.org/abs/1704.03155v2)
Usage: example_dnn_text_detection.exe [params]
-h, --help
Print help message.
--height (value:320)
Preprocess input image by resizing to a specific height. It should be multiple by 32.
-i, --input
Path to input image or video file. Skip this argument to capture frames from a camera.
-m, --model
Path to a binary .pb file contains trained network.
--nms (value:0.4)
Non-maximum suppression threshold.
--thr (value:0.5)
Confidence threshold.
--width (value:320)
Preprocess input image by resizing to a specific width. It should be multiple by 32.