Hello, I am attempting to reproduce your code on DFC2019 Track1, but the MAE of my test result is around 4.29. According to the performance of the comparison method presented in the paper "Depth2Elevation: Scale Modulation With Depth Anything Model for Single-View Remote Sensing Image Height Estimation", the MAE of Im2height on Track1 should be approximately 1.329. Therefore, I would like to ask if you have reproduced this method on Track1 and what the results were? I hope to figure out where I might have made a mistake.
Specifically, could you tell me how you processed the data input? Did you strictly follow the code in data.py? Is it unnecessary to perform normalization before input and denormalization after output?
Thank you very much for your time~
Hello, I am attempting to reproduce your code on DFC2019 Track1, but the MAE of my test result is around 4.29. According to the performance of the comparison method presented in the paper "Depth2Elevation: Scale Modulation With Depth Anything Model for Single-View Remote Sensing Image Height Estimation", the MAE of Im2height on Track1 should be approximately 1.329. Therefore, I would like to ask if you have reproduced this method on Track1 and what the results were? I hope to figure out where I might have made a mistake.
Specifically, could you tell me how you processed the data input? Did you strictly follow the code in data.py? Is it unnecessary to perform normalization before input and denormalization after output?
Thank you very much for your time~