The wR2, detection stage of the RPNet, may have good accuracy on CCPD. However, it can not locate well on other datasets, like UFPR-ALPR and AOLP. My personal opinion is that the license plates of CCPD are basically in the middle of the image, which may achieve satisfactory results. Is there any problem with the robustness of this Baseline to the LP datasets?
The wR2, detection stage of the RPNet, may have good accuracy on CCPD. However, it can not locate well on other datasets, like UFPR-ALPR and AOLP. My personal opinion is that the license plates of CCPD are basically in the middle of the image, which may achieve satisfactory results. Is there any problem with the robustness of this Baseline to the LP datasets?