Issue in regression and 3DCNN algorithm

Hi, I have developed a 3DCNN for regression of video images. The output measures RMSE and MAE. The problem is the value of RMSE and MAE is close to zero for 10 epoch and 5 fold cross validation. Others applied same database and different models. but most of the previous works results for RMSE and MAE is between 6 to 10. I think something is wrong in my code which measures RMSE and MAE very low. Any help appreciated. I was wondering if I implemented my algorithm based on 3DCNN and Regression in a correct way or not.

Double post from here. Without more information about your use case and your concern about a low loss it’s hard to help, so please add more information.

Thanks for your answer. It is an algorithm to inference depression level from AVEC database. I trained and evaluated the system with 10 epoch and applying the 5 fold cross validation. In each epoch the result for training and validation RMSE and MAE is close to zero for example: Train RMSE: 0.0167, Train MAE: 0.01 and Val RMSE: 0.0060, Val MAE: 0.01. The problem is other researchers in the same database with different deep learning algorithms achieved RMSE and MAE something between 6 to 10. Please let me if anything else I should share. I can share the code if needed. I was wondering if I implemented my algorithm based on 3DCNN and Regression in a correct way or not.