Superlee

The weight should multiply with each joint heatmeap not a scalar.

Why did you replace the noobject_loss?

3

Hey, I noticed that you replaced "f.nn.l2_loss(no_I * (p_C - C)) * self.noobject_scale" with "tf.nn.l2_loss(no_I * (p_C)) * self.noobject_scale". But as I know, the paper adopted the former one. Why...

Some places in loss_layer are different from the paper

1

 1. The loss equation describes the confidence loss of no object is " C-C' " , but in the code, the loss is "noobject_delta = noobject_mask * predict_scales" 2....

How can I summary a custom video?

1

As described, How can I test on a custom video and get the summary?

Line 538-539 is useless in dsst_tracker.hpp

1

In dss_tracker.hpp, [Line 538-539 ](https://github.com/klahaag/cf_tracking/blob/master/src/cf_libs/dsst/dsst_tracker.hpp#L538-L539)seems useless. Do I miss anything? ``` boundingBox &= Rect(0, 0, static_cast(image.cols), static_cast(image.rows)); boundingBox = tempBoundingBox; ```

As described above, why do you use the ESDR to smooth the results which is different from the paper. Thanks.

 As described above, why is the output channel 11?

请问给投影矩阵加上jitter的原理如何理解？

1

如题所述，您在代码里加上了随机偏移，我有一点不理解的是，4*4的投影矩阵，如果要在x，y方向加上偏移量，应该 是m03, m13, 为何这里是m02, m12？（2这个位置不应该是z分量吗） ``` var jitteredProjection = camera.projectionMatrix; jitteredProjection.m02 += (offset.x * 2 - 1) / camera.pixelWidth; jitteredProjection.m12 += (offset.x * 2 - 1) / camera.pixelHeight; ```

How to get the real depth value by the png image?

1

As described above.

Is this guiding method useful for light map baking( all the materials are diffuse)

1

Thanks for your great work. As described above, is this method useful for light map baking? In light map baking, the distance between revels is large due to the resolution...