Rahul Deora

Distance Transform

6

Hey, I do not understand how distance map is being used here and what the clicks variable is exactly supposed to represent: ``` def dt(a): return cv2.distanceTransform((a * 255).astype(np.uint8), cv2.DIST_L2,...

Infering on large images

1

As the model is trained on small patches, will it have trouble inferring on large 2000X2000 images as it has never see details of this larger size? Don't all matting...

Results

6

In Section 4.1 you say "The laplacian loss proposed by Hou et al. [13] gives a significant reduction in errors across all metrics. We note this network, training and loss...

The number of channels for offset and modulation usually has 27 channel output as kernal size is 3. But according to my understanding of paper, for each channel we a...

According to the paper and implementation the weights in the offset layer are initialised to 0 so the shift in x,y is 0 and modulation is 0.5. Doesnt it make...

DCNv2

1

How exactly is DCNv2 supported in this repo?

Questions about content and TF

2

Is the content exactly the same for the TF version as the original. I'm guessing if something cannot yet be done in TF it could have been skipped. There seems...

I am using google colab. How to you activate GPU usage for MCMC sampling? I am finding it very slow. And as building graphs is supposed to be faster than...

Logistic Example

3

In ch2 Logistic example, when we assume our data follows logistic, why then use Bernoulli after that? In previous examples we only had an assumption of the data distribution and...

A/B Testing Joint Explanation

1

The double_joint_log_prob in ch2 A/B testing is not explained at all. Why do we return ( rv_prob_A.log_prob(prob_A) + rv_prob_B.log_prob(prob_B) + tf.reduce_sum(rv_obs_A.log_prob(observations_A)) + tf.reduce_sum(rv_obs_B.log_prob(observations_B)) ) would'nt this corrospond to finding the...