Code to add object detection capability to ThisRobotAI
Note: Files for Tensorflow model are too large to upload. They need to be installed manually as per the instructions below.
Install This Robot AI software as per https://github.com/leematthewshome/robotAI_python3
Install necessary Python libraries
- sudo su -
- pip3 install imutils
- pip3 install opencv-python
- pip3 install tensorflow
Control-D to exit su mode
Copy the files and folders from this repository into robotAI/client/
Move cursor into the objectDetect folder
- cd objectDetect
Download model and unzip - note that the code is currently setup for the first model. You will need to edit code if you use a different model
Fast model but not most accurate
- wget http://download.tensorflow.org/models/object_detection/ssd_mobilenet_v1_coco_2017_11_17.tar.gz Trade off b/w speed and accuracy
- wget http://download.tensorflow.org/models/object_detection/faster_rcnn_inception_v2_coco_2018_01_28.tar.gz Most accurate, but slower
- wget http://download.tensorflow.org/models/object_detection/faster_rcnn_nas_coco_2018_01_28.tar.gz
Unzip into folder
- tar xvzf [filename]
Edit the RobotAI configuration to use camera as the motion detection mechanism.