pretrained model

[chamecall]

Hello guys, thanks for your work.

I have two questions to you:

1. is your provided "pretrained model" was trained by you on this dataset or by "pretrained" you mean pretrained on some other dataset like imagenet or smth?
2. do you have inference code?

[Vehicle-AHU]

1.Our pre-trained model is pre-trained on the Autobot1M dataset.
2.We directly apply the pre-trained model to train and infer with existing algorithms for various downstream tasks. You can download the corresponding code from the links provided below:
Vehicle Attribute Recognition: https://github.com/cxh0519/VTB
Vehicle Re-identification: https://github.com/heshuting555/TransReID
Vehicle Part Segmentation: https://github.com/fudan-zvg/SETR
Vehicle Fine-grained Recognition: https://github.com/TACJu/TransFG
Vehicle Detection: https://github.com/Event-AHU/VFM-Det

[chamecall]

1.Our pre-trained model is pre-trained on the Autobot1M dataset. 2.We directly apply the pre-trained model to train and infer with existing algorithms for various downstream tasks. You can download the corresponding code from the links provided below: Vehicle Attribute Recognition: https://github.com/cxh0519/VTB Vehicle Re-identification: https://github.com/heshuting555/TransReID Vehicle Part Segmentation: https://github.com/fudan-zvg/SETR Vehicle Fine-grained Recognition: https://github.com/TACJu/TransFG Vehicle Detection: https://github.com/Event-AHU/VFM-Det

1. Could you pls tell what exact task the model was pretrained on?
   Specifically I'm interested in vehicle type/color classification. Does the model fit for me or I need to finetune it on downstream task?
   Thanks

[Vehicle-AHU]

We conduct pre-training in a self-supervised manner, making it applicable to most vehicle perception tasks. The vehicle type and color classification, as well as attribute recognition tasks you mentioned, are consistent with our approach. You can use https://github.com/cxh0519/VTB to load our pre-trained model for fine-tuning.