This tutorial will help you use KITTI Vision Benchmark Dataset in Yonohub. KITTI Vision Benchmark is a complete dataset for computer vision research is the context of autonomous driving.
The tutorial goes through two main steps:
- Purchasing the KITTI Vision Benchmark Dataset and its player from YonoStore.
- Use the dataset player in YonoArc and connect it to your algorithms’ blocks. In this tutorial, we will use the Edge Detection block and YOLO Object Detection block as examples for image processing and computer vision algorithms respectively.
The following steps can be generalized and applied to all other datasets available on YonoStore: KITTI, Comm.ai, BDD, ApolloScape, nuScenes, Lyft, etc.
Check this to learn more about YonoStore.
Purchasing Assets from YonoStore
To purchase an asset from YonoStore:
- Navigate to YonoStore from the main view of Yonohub.
- Log in with your Yonohub account.
- Click on the Datasets icon.
- Type in the search bar “KITTI” and then press Enter.
- Choose the KITTI Raw Dataset from the list of search results.
- Choose Buyout from the Usage License drop-down menu. Click on Add to Cart.
- After adding the dataset to cart, it is essential to purchase the corresponding dataset player to interface with the dataset.
- Back to the search bar results, choose KITTI Raw Dataset Player.
- Choose Pay As You Go from the Usage License drop-down menu and then click on Add to Cart.
- Go to your cart by clicking on My cart button in the upper right corner of the page.
- Make sure you have all what you need. Finally, click on Proceed to checkout button.
- You will be asked for billing information before placing your order.
- After successfully placing your order, you can find your purchased items KITTI Raw Dataset and KITTI Raw Dataset Player in your YonoArc blocks.
KITTI Raw Dataset has the following output ports:
- Camera0 (gray image)
- Camera1 (gray image)
- Camera2 (colored image)
- Camera3 (colored image)
- Lidar (point cloud).
KITTI Raw Dataset Player has the following output ports:
- Original Image
- Segmented Image (the original image overlaid with segmentation data).
- Groundtruth Image (the groundtruth segmentation data)
Use KITTI Dataset in YonoArc Pipeline
- Click on the YonoArc icon from Yonohub’s Main View. You can follow this tutorial to get familiar with the YonoArc interface.
- Click the + button in the left upper corner, then click the Input section and choose KITTI Raw Dataset Player
- Configure the KITTI Raw Dataset Player block by clicking settings icon in the upper left corner of the block.
- Under the Properties tab, configure the paths for calibration data and synced and rectified data. Click on Browse -> YonoStoreDatasets -> rawdata -> road. Choose a zipped file for the calibration data as well as another zipped file for the Synced & Rectified data.
- Place the Edge Detection and YOLO Object Detection blocks from the Perception section in the canvas.
- Place Draw Boxes block from Others section.
- Place three Video Viewer blocks from the Dashboard section, rename each one of them from Settings in the upper left corner.
- Connect all the blocks as shown below. You can connect several blocks by select them and click “Ctrl + E“.
- Launch the pipeline and wait a while till you get all the blocks running, then open the dashboard by clicking on the Dashboard button in the bottom left corner.