Kevin Ke-Yun Lin

This repo does not support this function. One possible solution is to run instance segmentation using an existing model from other research works, and use the resultant instance segmentation map...

Thanks for your interest in our work! After company approval, we will release the training code and synthetic data.

Thank you very much for your interests in our work. Because the synthetic data and the 3D human models are copyrighted and belong to Microsoft, I am currently not authorized...

Thanks. I am checking this issue now. `eval_coco2014_multi_modes.py` loads a few modules including COCOAPI. It seems like this is an COCOAPI issue. Can you run the COCOAPI individually on your...

Thanks for your interests in our previous works. You can download our caffe models (SSDH trained on ILSVRC12 based on AlexNet or VGG16) in the following dropbox folder: https://www.dropbox.com/sh/t9ed398vdfdg7qk/AAAtk8u9ukOym502Ff_dUpDQa?dl=0

If you are asking the retrieval accuracy, the score is calculated with label. Label (or groundtruth) is used only for testing phase.

It's coarse-level search. In MNIST and CIFAR10 experiments, we evaluate the performance with only binary codes in order to fairly compare with other hashing approaches.

We first resize images to 256x256, and then center-crop 227x227 as network input. > On 2017年7月28日, at 上午2:17, 111Moderato wrote: > > name: "KevinNet_CIFAR10" layers { layer { name: "data"...

Yes. You can take a look the extension of this work here: https://arxiv.org/abs/1507.00101v2 Our workshop version can be seen as our proposed SSDH with the hyper parameters: alpha = 1,...

Oh.. I forgot that NUS-WIDE is a multi-label dataset.. Yes. You are right. Since this workshop version is trained with softmax loss, we cannot directly apply it on multi-label dataset....