Hello, when I used the provided navdp-cross-modal.ckpt to evaluate in the easy scenes with eval_imagegoal_wheeled.py, I found that the success rate was only around 30%. The evaluation results for the easy_2 scene are shown below (31%) is far from the data normally reported in the paper, and I'm really puzzled about what went wrong.
https://pan.baidu.com/s/18xpCIo9pvXCGiJanK8zzuQ?pwd=8e59
Hello, when I used the provided navdp-cross-modal.ckpt to evaluate in the easy scenes with eval_imagegoal_wheeled.py, I found that the success rate was only around 30%. The evaluation results for the easy_2 scene are shown below (31%) is far from the data normally reported in the paper, and I'm really puzzled about what went wrong.
https://pan.baidu.com/s/18xpCIo9pvXCGiJanK8zzuQ?pwd=8e59