Zhongyi Han

icon indicating a website link https://zhyhan.github.io/ icon indicating an email [email protected]

icon company KAUST icon location KSA icon location I am currently a Postdoctoral Fellow at KAUST.

Results 6 comments of Zhongyi Han

Total loss changes to Nan somehow

I have found that **logloss_tgt** may cause this problem because the **logloss_tgt** is easy to increasue to infinite when its input reaches to zero. Based on that, I have added...

keras 2.0 concat error in ssd.train ipython

Thank you @rykov8 , I changed the **get_output_shape** to **computer_output_shape** in **ssd_layers.py**, but it still appears same issues.

training voc2007

I got the same problems of convergence, the global step is about 8000 but the loss is about 6 and mAP is 0.026. I have no idea about it.

Question on different feature_extractor in CLIP

I have also encountered some difficulties in reproducing your results on the VLCS dataset. In Table 1 of the paper, the paper reported that DPLCLIP achieved an accuracy of 0.843,...

Question on different feature_extractor in CLIP

I attempt to choose the best performance (in early steps/epochs) to compute the average accuracy, and the performance is about 0.778 in the VLCS dataset. > > I have also...

Question on different feature_extractor in CLIP

I attempt to choose the best performance (in early steps/epochs) to compute the average accuracy, and the performance is about 0.778 in the VLCS dataset. > > I have also...