jinzhuxing / fingerprint_tf Goto Github PK

while training with sample dataset provided in dataset/ReadME.md file, i am getting this error.

Hy @JinZhuXing ,

Thanks for providing this repositories

Can we use this code to compare/verify fingerprint ?

I have setup this and run this code using pre-trained model.
I want to do fingerprint verification.

In this code after image pre-processing .npy files are generated in np_data folder.

And after running evaluation script this script not ask to provide a single image for which we need to compare/verify.

How can we customize this according to our use-case ?

Dataset's original folder not available in the dataset i downloaded from kaggle.
baidu isn't available in my area can you kindly share the link of of the dataset on google drive?

Hi,

i have started training on my custom dataset, but after every epoch, it is getting saved by the same name and getting rewritten after evry epoch. Also, i had given 200 epochs, a file has been created in model/result folder named fp160_160.h5 and it has once agin started from 0th epoch.

Epoch 00090: saving model to /home/vinay/github_repo/Fingerprint_TF/model/checkpoint/cp_160_160.ckpt

could you Please provide a way to save a model after every 5th or 10th epoch and naming convention for saved epoch?

Thank you

I have trained with 200 images..
python3 source/train.py --image_width 160 --image_height 160 --dataset_path dataset/original/ --check_point model/checkpoint/ --save_model 2  --save_model_path model/result/ --train_epoch 5  --eval_num 20

these are my parameters. The training never stops.  It's  been running for last one hour.
is there any bug in the code? 
I am not able to figure out what is happening. Please explain it to me why it is happening.Also suggest me how to proceed forward

Thanks

你好，你给出的百度网盘链接失效了，请问可以重新发一下吗？

Hi, I want to use the pretrained model derectly to train model. But after I changed the checkpoint folder and ran the 02_01.py, I got such error：Shapes (64,) and (32,) are incompatible， which comes from "model.load_weights(checkpoint_path)". I searched a lot but still confused. Please help me!!! Thank you!

in the dataset/ReadME.md file, for train_data real_data and np_data, 3 separate scripts are mentioned, so that the respective folders are created. But the scripts mentioned in the files are not present in this repository. Please add the scripts required for creating the dataset to the main branch.
Thanks in advance. :)

It seems the dataset you're sharing on Kaggle and GDrive has only 10 fingers (800 images) but the pretrained model you provided are trained with much larger dataset. Could you please give a pointer to those data sources?