Xiongkun Linghu

The inference process is very slow

1

I run the default scripts, however, 100 generated samples take about 15 hours, is there any method to accelarate the process?

All the results are lower than paper with the default settings and feature pyramid 5,2,1

1

I have tried FCN pyramid structure 5,2,1 in the paper, but the accuracy on miniImagenenet is still 65.5%. Besides, acc on tiered imagenet is 70.6%(2% lower), 72.8%(2% lower) on CIFAR-FS,...

Can't achieve the accuracy in the papar

5

I pretrained the model and then used deep emd with the default setting, but the 5-way-1-shot accuracy was just 65.5%, which was 1% lower than the paper.

Fail to train in mini-Imagenet

5

I use the edl loss to train in mini-imagenet dataset with 64 classes, but the loss can't converge and the accuracy is very low.

About the perceptual distance

1

I want to caculate the perceptual distance for the specific input images, could you please provide the detailed impletation of this part?

The work is interesting. I want to train my model with your datasets. Could you please provide more detailed description of the datasets used in the Table 1 in the...

I want to generate samples interpolated by 2 images, is it covered by the codebase?

How to get grounding data?

1

Thanks for sharing the work. I notice that the model can output coordinates of the 3D bounding boxes throught numerical values. How to access this data related to 3D grounding...

Explanation for data format and issues about data generation

2

Thanks for your insteresting work. I visualize the grounded scene caption data and notice there is a key called 'all_phrases_positions'. What does it mean? I guess the numerical values represent...

Encoding referent tokens

4

I notice referent tokens are interleaved in the output. Can multiple referent tokens appear in a single text prompt, such as "Describe the table and the chair ."?