search

yg33717 / DSH_tensorflow

implemement of DEEP SUPERVISED HASHING FOR FAST IMAGE RETRIEVAL_CVPR2016
105 stars 37 forks source link

About validation #7

Open soon-will opened 5 years ago

soon-will commented 5 years ago

Thank you for your effort. While training, you save the model after 10 epochs, but it seems that you don't set val_set to val the model. Then how do you ensure the model saved is the best model?

yuyifan1991 commented 5 years ago

@soon-will Do you know what does the result.txt mean? How to describ the results of the image-retrieval?

soon-will commented 5 years ago

@yuyifan1991 Sorry that, I don't run the demo, but it seems that result.txt is consist of hash code and label. You can run the demo.py to confirm that.

As for the way to describ the results of similarity image retrieval (deep hashing), mAP/PR curves/P@H=2 is a standard evaluation protocol.

yuyifan1991 commented 5 years ago

@soon-will Thank you first, and now I just want to know how to compute the similarities between the two pictures. I have got the mAP by running the caculate_mAP.py .