About datasets with images

[Bottle101]

Note: The rosbags with images are within the corresponding zipped files on the drive. For example, 20230403_hkustgz_vegetation_sequence01_r3live_semantics_framecam00.bag is within hkustgz_vegetation_sequence01.tar.7z

---

Hi botao,

Thanks for your interest in our work.

These rosbags should provide images:

20220226_campus_road_day_r3live_semantics_framecam00.bag 20230403_hkustgz_vegetation_sequence00_r3live_semantics_framecam00.bag

20230403_hkustgz_vegetation_sequence01_r3live_semantics_framecam00.bag

Please subscribe the topic called: `` /r3live/rectified_img/compressed’’ (which is already undistorted) (see the figure). We only publish keyframe images (with a rotation and translation threshold) for segmentation. This strategy can help to reduce the duplicate computation of the same scene.

If you want images from data with the name such as r3live_result_20220216_canteen_day.bag’’, you need to follow these steps: 1) download the raw rosbag called 20220216_canteen_day.bag’’ from the fusionportable website: https://fusionportable.github.io/dataset/fusionportable, 2) do image undistortion using calibration files: https://drive.google.com/drive/folders/123dzzRnMxVD0xnprB-AOrbiaB76-YSdv?usp=drive_link , and then 3) record all distorted image and merge these images with ``r3live_result_20220216_canteen_day.bag’’

Please tell me if you have any other questions.

BTW, could you please create an issue in the github?

Thanks.

Yours sincerely,

Jianhao Jiao.

---

From: Botao He <[email protected]>
Date: Wednesday, October 9, 2024 at 00:11
To: [email protected] <[email protected]>
Subject: Question regarding the Semantic Mapping dataset

Hi Dr. Jiao,

I am Botao He, a PhD student from the University of Maryland, I am writing to ask a question regarding your recent work about semantic mapping dataset titled "Real-Time Metric-Semantic Mapping for Autonomous Navigation in Outdoor Environments".

Could you kindly tell me which dataset contains Lidar scans as well as images? I downloaded several datasets and only found lidar scans, but I would like to do semantic segmentation on images as well. In the paper I saw cameras are also mounted, is there any dataset contains images?

Looking forward to hearing from you,

Botao