Video demonstration: https://youtu.be/6v8h8LPFRls

Description

Tensorflow model adapted for DSCVL.

https://github.com/tensorflow/models/tree/master/research/object_detection

Dependencies

Tested, developed, and executed on Windows 10, build 1803

1. Python 3.5.2 - https://www.python.org/downloads/release/python-352/
2. Tensorflow-gpu to run visualization recommended
   • pip3 install --upgrade tensorflow-gpu
3. RPlidar A1 - https://www.slamtec.com/en/lidar/a1
   • pip3 install rplidar
4. Matplotlib
   • pip3 install matplotlib
5. Pillow
   • pip3 install pillow
6. OpenCV 3.4.0.12
   • pip3 install opencv-python==3.4.0.12
7. CP210x USB to UART Bridge VCP Driver (If on Windows 10, build 1803, please see below)
   • https://www.silabs.com/products/development-tools/software/usb-to-uart-bridge-vcp-drivers

Tensorflow-gpu

3. Nvidia CUDA Toolkit 9.0.176
4. Nvidia CUDnn v7.0.4 for CUDA 9.0

Windows 10, build 1803 Driver Fix

The CP210x USB to UART Bridge VCP Driver, required for RPlidar A1, does not meet Windows driver INF requirements and will not install.

Read more

Instructions to enable unsigned driver installation

My modified driver download - https://drive.google.com/file/d/1jKxpAml_JoFPP2b4fXIjJWZ2_udghgj9/view?usp=sharing

Run Visualization

1. Install all dependencies
2. Clone repository
   • git clone https://github.com/DSCVL/DSCVL-Vision.git
3. In file lam_helper.py, replace line 71 PORT_NAME = 'COM15' with the COM port that the RPlidar A1 is connected to.
4. Ensure web cam and RPlidar are plugged in, execute with main.py
   • python main.py