Why the training effect is good on my own data, but the testing effect is poor？

[fanzz1208]

Training progress: 100%|__________________________________________________________| 30000/30000 [1:58:29<00:00, 4.22it/s, Loss=0.0295401]
2024-03-14 03:49:43,489 - INFO:
[ITER 30000] Evaluating test: L1 0.2103012849887212 PSNR 11.615882555643717
2024-03-14 03:50:12,150 - INFO:
[ITER 30000] Evaluating train: L1 0.01723341103643179 PSNR 32.359920501708984

[inspirelt]

According to your previous issue #29 (comment), I guess that you pose is very poor, making the learned gaussians fail to generalize to novel views.