Jinkun Cao

@JLJ19 Can you try to make your contribution wrapped under the `deploy` folder and make a PR?

We recently have an improved version of combining OC-SORT + Deep SORT, which we name as Deep OC-SORT. You can check the details [here](https://github.com/GerardMaggiolino/Deep-OC-SORT).

For the best, you should also add clamp in the focal_loss in case the sigmoid produce `inf` value: ``` def focal_loss_alt(self, x, y): '''Focal loss alternative. Args: x: (tensor) sized...

Oh, yes, you are right! That is a mistake. Thanks for finding this issue!!!

Did you also run the linear interpolation stage? The results on MOT17 and MOT20 are by default after the post-processing of linear interpolation as in previous methods.

Yes, I'm also looking forward to that, or at least the raw experiments logs as other datasets.

@neonkitchen I think you should resize the saved image array. Because the default implementation of [Shapes3D dataset ](https://github.com/google-research/disentanglement_lib/blob/86a644d4ed35c771560dc3360756363d35477357/disentanglement_lib/data/ground_truth/shapes3d.py#L58)requires to reshape the loaded image when using.

Hi @muelea Thanks for the update. But the Auxiliary data for Hi4D and CHI3D seems still missing as the links provided [here](https://github.com/muelea/buddi/blob/main/documentation/DATA.md) are not valid.

Thanks for the prompt response! Sure, this is a point cloud I sample from my dataset: This is the reconstructed by using the function `VAE.recont(points)` [link](https://github.com/nv-tlabs/LION/blob/155a22f5c9f5ff4b2d15aed4e86fbdd8b4bf7ba1/models/vae_adain.py#L137) ![Screenshot 2024-03-26 at 18...

Also, I didn't use the `mixing_component` function for my training. Will this matter that lot?