Open Tianlu-Zhang opened 4 years ago
Hi,
Yes, the visualization analysis shows on the paper is evaluated on the testing dataset, very similar distribution with the training data (i.e. KAIST dataset), which explicitly help us select a better translation model to transfer large-scale RGB tracking datasets to TIR modality. The models fine-tuned with generated training data does improve the TIR tracking performance.
Thanks for your attention.
/Lichao
Hi, thank you for sharing. I downloaded your weight file. But I encountered a problem using https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix. CycleGAN can generate thermal infrared images, but pix2pix reports an error. I enter the command in the terminal as: python test.py --dataroot ./datasets/visible/ --direction AtoB --model test --name net_G_pix2pix. But the error message is AttributeError: 'Sequential' object has no attribute 'model'.
Yes, opening your generated TIR dataset is beneficial for community research. For our convenience, could you upload and share your data?
I tried to use your way to generate thermal images,but I cannot get the results like the results in the paper. So do you have a plan to open the generated dataset,thanks for your answers.
And do you use this code to generate images:https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix