How to make the reconstruction more accurate?

[Wudierer]

Hi! I tried to run the code on my computer. I didn't make any changes and the parameters were set as the sample in the README. And the data set was generated by synthetic_data.py. But the outline of the reconstructed signal is not clear. For example, single-rshort after reconstruction sometimes looks like single-cont. I tried to set the number of training epochs to 1000, but I got the same result.
This is the result of 1000 epochs of training.

Here is my requirements:
python 3.6.5 conda 4.5.4
tensorflow 1.12.0
Intel(R) Core(TM) i7-10700
NVIDIA GeForce GTX 1660 SUPER

I was wondering if you would give me some advice to make the reconstruction result more accurate.

[zeroXzero]

@Wudierer You can try to improve the AE decoder and encoder CNN layers to better capture input variations. Also try with more intermediate features.