RaTrack

RaTrack copied to clipboard

Hello, thanks for sharing this amazing work. I have some questions about the trained weight and evaluation results.

1. Can I use the trained weights from checkpoint/track4d_radar/models/model.last.t7 to directly infer and evaluate on the VOD dataset? When I used the weight from it for direct inference on the VOD dataset, I noticed that the results of motion segmentation and clustering (after removing clusters with less than 5 points) were quite noisy in some frames.

Frame around 410

Frame around 4729

2. Would it be possible to share the code that can visualize the result in the same way as shown in the demo video? (Include the tracking path and filter out the moving objects less than 5 points).