Tim Meinhardt
I just checked the demo code and it works on my side. Did u install the correct PyTorch etc. versions? Can u run the regular MOT17 evaluation of our code...
The variable `mask` of the NestedTensor which is also passed to the encoder does not refer to any output segmentation mask. This is a mask which is used to stack...
The question is not really related to `deformable=False` and `args.Tracking=False`, right? We keep and stack the intermediate outputs of the decoder for the auxilliary losses we compute at each layer....
The model uses the output of the last layer during testing/inference. The others are only for training.
They are part of DETR and Deformable DETR.
Please read the paper and its related work to fully understand how our method is working.
Hello, changing the backbone to something different might involve quite some changes to the code which depends a lot on the backbone you want to integrate. Such an extension goes...
You could reduce the input resolution or if you are interested in reducing overall sequence processing time, you could forward the video with less frames.
There is a difference between `image_id` and `frame_id`. Please check the `generate_coco_from_mot.py` to understand each field. The latter does not exist in the original COCO annotations and is related to...
Hello, I am trying to follow yoru explanation but what do u mean by `in an image of 55 frames`? In a MOT dataset the track ids will not align...