- 1 Project Overview
- 2 Team Members
- 3 Design
- 4 Lane Finding with OpenCV
- 5 Project
Our project was to integrate OpenCV into donkeycar in order to try and improve training time and robustness of the model. We used OpenCV to pre-process images in order to extract edges and highlight the white and yellow lanes. We wanted to see if using those images as a combined 3-D image (edges, white lane, yellow lane) would improve training time and robustness of the model compared to the original donkeycar model which trained using RGB images.
- Cyrus Shen - ECE
- Chenfeng Wu - ECE
- Maximilian Stadler - CSE (UCSD Extension)
- Isabella Franco - MAE
Aspect Ratios and Cropping
The original configuration suggested the recording of images with the size of 160x120 pixels resulting in a 4 by 3 aspect ratio and doubling the resolution for the outdoor setting, i.e. to 320x240. The used camera is itself a 2 MP camera being capable of recording FHD images. Thus its native resolution is 16 by 9. To allow for easy comparison, we decided to record indoor and outdoor images at the same resolution. We furthermore tried different aspect ratios to find how much information, especially in the corners, is lost due to cropping. We modified the framework such that only the height and the aspect ratio are specified in the configuration files. The image width is derived. We tested 1 by 1, 4 by 3 and 16 by 9. We concluded that the native wide angle resolution gives us the most information in the corners, which seemed to be prone to errors in the beginning anyway.
The Donkeycar-Framework is coded in a relatively rigid fashion. Many hard-coded sections and in-line definitions make any customization hard. Our goal was to compare the original models with models using visual primitives as priors, namely edges and lane segmentations. While these primitives can be easily computed using OpenCV, coding had to be done carefully to keep the code modular and allow an easy comparison of the original version without any code changes and the OpenCV-based models. Furthermore, to allow flexible modifications, our code was kept parametric which resulted in several additions to config.py allowing for customizable preprocessing.
TODO add a general overview of code-changes
TODO add config allowing for customization
Lane Finding with OpenCV
HSL Colors for Lane Detection
In order to single out the white and yellow lanes from the rest of the image, we used HSL color-space to threshold the image for the colors white and yellow. The RGB image taken by the camera was converted to HSL using OpenCV's COLOR_RGB2HLS and then thresholded so that only white and yellow colors in the image were kept.
- insert white lane image
- insert yellow lane image
- insert HSL info images
Canny Edge Detection
To outline the lanes, we used OpenCV's canny function which shows edges in an image.
- insert canny edge image
Combined 3-D image
We combined the white, yellow, and canny edge images into a 3-D image and trained on that instead of the original RGB image.
- insert new 3D image
In order to compare our OpenCV model to the current donkeycar model, we trained both models on the same images. 6 separate models were trained, 3 for an indoor track and 3 for an outdoor track. The 3 indoor track models were trained with 1000, 5000, and 10000 images of the indoor track with the car going clockwise. The same was done for the outdoor track model except the images were taken from the outdoor track.
We tested each model based on the following criteria:
- Training time
- Completion of X clockwise laps on the indoor track
- Completion of X counterclockwise lap on the indoor track
- Completion of X clockwise laps on the outdoor track
- Completion of X counterclockwise lap on the outdoor track
- Able to drive when bright or dark outside?
Indoor Track Model Results
Outdoor Track Model Results