RGB-T epoch 60 inference weights

RGB-T inference weights from training epoch 60, for experiments/lifttrack/rgbt_lifttrack.yaml.

• Asset: LiftTrack_RGBT_ep0060.pth.tar (approximately 354 MiB).
• Contains all 262 model state entries, identical to the original training checkpoint. Optimizer, scheduler, RNG state, and local settings are omitted; this asset is for inference, not exact training resume.
• Validation: safe weights-only deserialization, strict model loading, and CUDA forward outputs identical to the original checkpoint on the same random inputs.
• Correspondence with the final paper results has not yet been verified. This release makes no benchmark-performance claim. RGB-D and RGB-E weights are not included.

Download both the weight file and SHA256SUMS.txt to the same folder, then verify with sha256sum -c SHA256SUMS.txt.

Example from the repository root, after downloading the weight file into pretrained/ and configuring the dataset path:

python RGBT_workspace/test_rgbt_mgpus.py --dataset_name LasHeR --yaml_name rgbt_lifttrack --checkpoint pretrained/LiftTrack_RGBT_ep0060.pth.tar --threads 1 --num_gpus 1

SHA-256: 8a1d4b79a78ce7da50d29411f72e921139fa0175e7870b7106e84c2f685807ea

Source code, RGB-T weights, and checksums are archived on Zenodo: 10.5281/zenodo.23483732.