Skip to content

Latest commit

 

History

History
190 lines (186 loc) · 9.19 KB

OBJECT_DETECTION.md

File metadata and controls

190 lines (186 loc) · 9.19 KB
Raw

Object Detection

Launching

OpenSource Version

mobilenet-ssd

  • download and convert a trained model to produce an optimized Intermediate Representation (IR) of the model 
    cd /opt/openvino_toolkit/open_model_zoo/tools/downloader
python3 ./downloader.py --name mobilenet-ssd
#FP32 precision model
sudo python3 /opt/openvino_toolkit/dldt/model-optimizer/mo.py --input_model /opt/openvino_toolkit/open_model_zoo/tools/downloader/public/mobilenet-ssd/mobilenet-ssd.caffemodel --output_dir /opt/openvino_toolkit/models/object_detection/mobilenet-ssd/caffe/output/FP32 --mean_values [127.5,127.5,127.5] --scale_values [127.5]
#FP16 precision model
sudo python3 /opt/openvino_toolkit/dldt/model-optimizer/mo.py --input_model /opt/openvino_toolkit/open_model_zoo/tools/downloader/public/mobilenet-ssd/mobilenet-ssd.caffemodel --output_dir /opt/openvino_toolkit/models/object_detection/mobilenet-ssd/caffe/output/FP16 --data_type=FP16 --mean_values [127.5,127.5,127.5] --scale_values [127.5]
  • copy label files (excute once)
    sudo cp /opt/openvino_toolkit/ros_openvino_toolkit/data/labels/object_detection/mobilenet-ssd.labels /opt/openvino_toolkit/models/object_detection/mobilenet-ssd/caffe/output/FP32
sudo cp /opt/openvino_toolkit/ros_openvino_toolkit/data/labels/object_detection/mobilenet-ssd.labels /opt/openvino_toolkit/models/object_detection/mobilenet-ssd/caffe/output/FP16
  • run object detection sample code input from RealSenseCamera.(connect Intel® Neural Compute Stick 2) 
    roslaunch vino_launch pipeline_object.launch
  • run object detection sample code input from RealSenseCameraTopic.(connect Intel® Neural Compute Stick 2) 
    roslaunch vino_launch pipeline_object_topic.launch

YOLOv2-voc

  • Darkflow to protobuf(.pb)
    • install darkflow
      • install prerequsites
      pip3 install tensorflow opencv-python numpy networkx cython
      • Get darkflow and YOLO-OpenVINO
      mkdir -p ~/code && cd ~/code
git clone https://github.com/thtrieu/darkflow
git clone https://github.com/chaoli2/YOLO-OpenVINO
sudo ln -sf ~/code/darkflow /opt/openvino_toolkit/
      • modify the line self.offset = 16 in the ./darkflow/utils/loader.py file and replace with self.offset = 20
      • Install darkflow
      cd ~/code/darkflow
pip3 install .
    • Copy voc.names in YOLO-OpenVINO/common to labels.txt in darkflow. 
      cp ~/code/YOLO-OpenVINO/common/voc.names ~/code/darkflow/labels.txt
    • Get yolov2 weights and cfg 
      cd ~/code/darkflow
mkdir -p models
cd models
wget -c https://pjreddie.com/media/files/yolov2-voc.weights
wget https://raw.githubusercontent.com/pjreddie/darknet/master/cfg/yolov2-voc.cfg
    • Run convert script 
      cd ~/code/darkflow
flow --model models/yolov2-voc.cfg --load models/yolov2-voc.weights --savepb
  • Convert YOLOv2-voc TensorFlow Model to the optimized Intermediate Representation (IR) of model 
    cd ~/code/darkflow
# FP32 precision model
sudo python3 /opt/openvino_toolkit/dldt/model-optimizer/mo_tf.py \
--input_model built_graph/yolov2-voc.pb \
--batch 1 \
--tensorflow_use_custom_operations_config /opt/openvino_toolkit/dldt/model-optimizer/extensions/front/tf/yolo_v2_voc.json \
--data_type FP32 \
--output_dir /opt/openvino_toolkit/models/object_detection/YOLOv2-voc/tf/output/FP32
# FP16 precision model
sudo python3 /opt/openvino_toolkit/dldt/model-optimizer/mo_tf.py \
--input_model built_graph/yolov2-voc.pb \
--batch 1 \
--tensorflow_use_custom_operations_config /opt/openvino_toolkit/dldt/model-optimizer/extensions/front/tf/yolo_v2_voc.json \
--data_type FP16 \
--output_dir /opt/openvino_toolkit/models/object_detection/YOLOv2-voc/tf/output/FP16
  • copy label files (excute once)
    sudo cp /opt/openvino_toolkit/ros_openvino_toolkit/data/labels/object_detection/yolov2-voc.labels /opt/openvino_toolkit/models/object_detection/YOLOv2-voc/tf/output/FP32  
sudo cp /opt/openvino_toolkit/ros_openvino_toolkit/data/labels/object_detection/yolov2-voc.labels /opt/openvino_toolkit/models/object_detection/YOLOv2-voc/tf/output/FP16
  • run object detection sample code input from RealSenseCamera.(connect Intel® Neural Compute Stick 2) 
    roslaunch vino_launch pipeline_object_yolo.launch
  • run object detection sample code input from RealSenseCameraTopic.(connect Intel® Neural Compute Stick 2) 
    roslaunch vino_launch pipeline_object_yolo_topic.launch

Binary Version

mobilenet-ssd

  • download and convert a trained model to produce an optimized Intermediate Representation (IR) of the model 
    cd /opt/intel/openvino/deployment_tools/tools/model_downloader
sudo python3 ./downloader.py --name mobilenet-ssd
#FP32 precision model
sudo python3 /opt/intel/openvino/deployment_tools/model_optimizer/mo.py --input_model /opt/intel/openvino/deployment_tools/open_model_zoo/tools/downloader/public/mobilenet-ssd/mobilenet-ssd.caffemodel --output_dir /opt/openvino_toolkit/models/object_detection/mobilenet-ssd/caffe/output/FP32 --mean_values [127.5,127.5,127.5] --scale_values [127.5]
#FP16 precision model
sudo python3 /opt/intel/openvino/deployment_tools/model_optimizer/mo.py --input_model /opt/intel/openvino/deployment_tools/open_model_zoo/tools/downloader/public/mobilenet-ssd/mobilenet-ssd.caffemodel --output_dir /opt/openvino_toolkit/models/object_detection/mobilenet-ssd/caffe/output/FP16 --data_type=FP16 --mean_values [127.5,127.5,127.5] --scale_values [127.5]
  • copy label files (excute once)
    sudo cp /opt/openvino_toolkit/ros_openvino_toolkit/data/labels/object_detection/mobilenet-ssd.labels /opt/openvino_toolkit/models/object_detection/mobilenet-ssd/caffe/output/FP32
sudo cp /opt/openvino_toolkit/ros_openvino_toolkit/data/labels/object_detection/mobilenet-ssd.labels /opt/openvino_toolkit/models/object_detection/mobilenet-ssd/caffe/output/FP16
  • run object detection sample code input from RealSenseCamera.(connect Intel® Neural Compute Stick 2) 
    roslaunch vino_launch pipeline_object.launch
  • run object detection sample code input from RealSenseCameraTopic.(connect Intel® Neural Compute Stick 2) 
    roslaunch vino_launch pipeline_object_topic.launch

    YOLOv2-voc

  • Darkflow to protobuf(.pb)
    • install darkflow
      • install prerequsites
      pip3 install tensorflow opencv-python numpy networkx cython
      • Get darkflow and YOLO-OpenVINO
      mkdir -p ~/code && cd ~/code
git clone https://github.com/thtrieu/darkflow
git clone https://github.com/chaoli2/YOLO-OpenVINO
sudo ln -sf ~/code/darkflow /opt/openvino_toolkit/
      • modify the line self.offset = 16 in the ./darkflow/utils/loader.py file and replace with self.offset = 20
      • Install darkflow
      cd ~/code/darkflow
pip3 install .
    • Copy voc.names in YOLO-OpenVINO/common to labels.txt in darkflow. 
      cp ~/code/YOLO-OpenVINO/common/voc.names ~/code/darkflow/labels.txt
    • Get yolov2 weights and cfg 
      cd ~/code/darkflow
mkdir -p models
cd models
wget -c https://pjreddie.com/media/files/yolov2-voc.weights
wget https://raw.githubusercontent.com/pjreddie/darknet/master/cfg/yolov2-voc.cfg
    • Run convert script 
      cd ~/code/darkflow
flow --model models/yolov2-voc.cfg --load models/yolov2-voc.weights --savepb
  • Convert YOLOv2-voc TensorFlow Model to the optimized Intermediate Representation (IR) of model 
    cd ~/code/darkflow
# FP32 precision model
sudo python3 /opt/intel/openvino/deployment_tools/model_optimizer/mo_tf.py \
--input_model built_graph/yolov2-voc.pb \
--batch 1 \
--tensorflow_use_custom_operations_config /opt/intel/openvino/deployment_tools/model_optimizer/extensions/front/tf/yolo_v2_voc.json \
--data_type FP32 \
--output_dir /opt/openvino_toolkit/models/object_detection/YOLOv2-voc/tf/output/FP32
# FP16 precision model
sudo python3 /opt/intel/openvino/deployment_tools/model_optimizer/mo_tf.py \
--input_model built_graph/yolov2-voc.pb \
--batch 1 \
--tensorflow_use_custom_operations_config /opt/intel/openvino/deployment_tools/model_optimizer/extensions/front/tf/yolo_v2_voc.json \
--data_type FP16 \
--output_dir /opt/openvino_toolkit/models/object_detection/YOLOv2-voc/tf/output/FP16
  • copy label files (excute once)
    sudo cp /opt/openvino_toolkit/ros_openvino_toolkit/data/labels/object_detection/yolov2-voc.labels /opt/openvino_toolkit/models/object_detection/YOLOv2-voc/tf/output/FP32
sudo cp /opt/openvino_toolkit/ros_openvino_toolkit/data/labels/object_detection/yolov2-voc.labels /opt/openvino_toolkit/models/object_detection/YOLOv2-voc/tf/output/FP16
  • run object detection sample code input from RealSenseCamera.(connect Intel® Neural Compute Stick 2) 
    roslaunch vino_launch pipeline_object_yolo.launch
  • run object detection sample code input from RealSenseCameraTopic.(connect Intel® Neural Compute Stick 2) 
    roslaunch vino_launch pipeline_object_yolo_topic.launch