can't get good result

[dashidhy]

Hi,

I'm a researcher who is also working on automatic particle picking technique. I have read your paper and follow your steps to test your code, but I can't obtain the same result as your said in the paper. The picture below is one of the TRPV1 micrograph I downloaded from EMPIAR-10005 (one of the same dataset as you used in your paper) and I used it to test your demo model. The blue circles are human-picked particles, while the red ones are picked by the demo model. As you can see the result is not ideal. I wonder if you have done some preprocessing before using the micrographs? And I will appreciate it if you could provide some of your original data for training the CNN model and testing the results. Thanks!

[nejyeah]

Hi, Sorry for replying late.
When training the model, all the data were already lowpassed by EMAN2 script, set the threshold as 0.01 maybe. Besides, we noticed that the first 100 micrographs in TRPV1 dataset are not clear for human to see, we choose the later 400 hundred micrographs to train and test, which are much better.
Besides, I have no original data in hand. I have graduated and work in company now.
If you meet any problems, please let me know, thanks.

[dashidhy]

@nejyeah Thanks a lot! I will have a try.