the result is not good

[QueenJuliaZxx]

Hi,you,I have trained the model on train for a long time ,but the detect result is too bad,could you tell me why does it happened~thank you

[StasDeep]

I'm not sure whether I'm doing everything right, but after 150K batches I have very poor results, too. P.S. Running on Python 3.6 and TensorFlow 1.3.

UPD. Decreased learning rate and executed 220K batches, but still no success. It's worth to mention, that predictions are the same for all test examples after both 1K and 220K batches.

[wenting521]

Can you share your code? I did get an error while executing train.py. (1、AttributeError: Can't pickle local object 'mpgen..main'； 2、EOFError: Ran out of input.) Do you have encountered such problems and hope to get your answers. Thank you very much!

[mazcallu]

I get a detection rate in videos that vary from 50% to 90%, I fine tuned the algorithm for detecting european plates in specific conditions, the detection can also be improved by modifying the highest probability candidate code to a majority voting solution which makes it more robust to noise influence, but the whole algorithm has limitations in real world scenarios due to the synthetic nature of the dataset, imo

[Rafi332]

I get a detection rate in videos that vary from 50% to 90%, I fine tuned the algorithm for detecting european plates in specific conditions, the detection can also be improved by modifying the highest probability candidate code to a majority voting solution which makes it more robust to noise influence, but the whole algorithm has limitations in real world scenarios due to the synthetic nature of the dataset, imo

can you plz share your code with me?/ my email id : [email protected]