yaboo-oyabu / carnd-capstone Goto Github PK

A README file in .txt, .md, or .pdf format. This file should contain the full names and Udacity account emails of your team members. For additional resources on creating READMEs or using Markdown, see here or here.

Collect images to train tl_detector for the virtual environment.

In order to help you acquire an accurate ground truth data source for the traffic light classifier, the Udacity simulator publishes the current color state of all traffic lights in the simulator to the /vehicle/traffic_lights topic in addition to the light location. This state can be used to generate classified images or subbed into your solution to help you work on another single component of the node. The state component of the topic won't be available when running your solution in real life so don't rely on it in the final submission. However, you can still reference this topic in real life to get the 3D world position of the traffic light.

Note that you need to upgrade pillow to 6.2.2 to get messages from /image_color topic.

After your team has confirmed that traffic light detection is working in the simulator, you should test it out using ROS bags that were recorded at the test site: Traffic Light Detection Test Video.

https://drive.google.com/file/d/1aanZVdWSWasBmz1J-hcOz39N6aVEeHlf/view

List up action items as GitHub Issues so that we won't reinvent the wheel inside the team.

We may not need to train traffic light detection model for real world because there are many publicly available object detection models that can detect traffic lights. https://github.com/tensorflow/models/tree/master/research/object_detection

This video will be a good source: https://s3-us-west-1.amazonaws.com/udacity-selfdrivingcar/advanced_deep_learning/driving.mp4

Submit the project by following the instruction in https://classroom.udacity.com/nanodegrees/nd013/parts/01a340a5-39b5-4202-9f89-d96de8cf17be/modules/702b3c5a-b896-4cca-8a64-dfe0daf09449/lessons/e43b2e6d-6def-4d3a-b332-7a58b847bfa4/project

@yaboo-oyabu changed the default model from mobilenet_v2 based model to inception_v2 based model. But it seems this makes inference speed slower than realtime. The default model should be changed to a more lightweight model.

Integrate the following ML models with tl_detector.

• object detection: ssd_inception_v2_retrained_2806
• classification: tl_classify_mixed_extracted.h5

Note that the traffic light class becomes 1 (one). Previously it was 10.

Let's check dependencies on external libraries for core code which will be used in a test course.

The project that is evaluated on the vehicle should not have any external libraries that aren't included in the starter repo. Your team should be aware of the hardware specifications of the Udacity vehicle when designing your solution.

The CarND-Capstone/ros/src/waypoint_loader/launch/waypoint_loader.launch file is set up to load the waypoints for the first track. To test using the second track, you will need to change

<param name="path" value="$(find styx)../../../data/wp_yaw_const.csv" />

to use the churchlot_with_cars.csv as follows:

<param name="path" value="$(find styx)../../../data/churchlot_with_cars.csv"/>

Note that the second track does not send any camera data.