dataset

[Shadow-417]

Great work！ I have a few questions to clarify some details, and would really appreciate your help: 1. Is the 3M of CRSD referring to the features extracted using different encoders on the RS5M dataset? 2. Is the original dataset obtained from https://huggingface.co/datasets/Zilun/RS5M/tree/main? 3. Where can the original dataset corresponding to the 1M of LRSD be obtained? 4. Which method is used to obtain the vector database for the final CRSD and LRSD? Thank you for your time and help!

[zilunzhang]

Great work！ I have a few questions to clarify some details, and would really appreciate your help: 1. Is the 3M of CRSD referring to the features extracted using different encoders on the RS5M dataset? 2. Is the original dataset obtained from https://huggingface.co/datasets/Zilun/RS5M/tree/main? 3. Where can the original dataset corresponding to the 1M of LRSD be obtained? 4. Which method is used to obtain the vector database for the final CRSD and LRSD? Thank you for your time and help!

1. Yes
2. Yes, for CRSD.
3. You can download the datasets by referencing Table 1. I just cropped the objects using annotations.

4. I am not sure if I understand the question correctly. "Chroma is chosen as our vector database. Label and caption information from LRSD and CRSD are encoded using Sentence-Bert (“all-MiniLM-L6-v2” model) and stored
   in Chroma."