| --- |
| license: mit |
| --- |
| |
| Two versions of the BLT-VS model (see: https://github.com/KietzmannLab/BLT-VS) are available. |
|
|
| Model 1: blt_vs_slt_111_biounroll_1_t_12_readout_multi_dataset_ecoset_num_1 (trained on Imagenet-1k; ILSVRC) |
| |
| timesteps=12, |
| num_classes=1000, |
| add_feats=100, |
| lateral_connections=True, |
| topdown_connections=True, |
| skip_connections=True, |
| bio_unroll=True, |
| image_size=224, |
| hook_type='None', |
| readout_type='multi' |
|
|
| Model 2: blt_vs_slt_111_biounroll_0_t_6_readout_multi_dataset_imagenet_num_1 (trained on Ecoset-565 - https://huggingface.co/datasets/kietzmannlab/ecoset) |
| |
| timesteps=6, |
| num_classes=565, |
| add_feats=100, |
| lateral_connections=True, |
| topdown_connections=True, |
| skip_connections=True, |
| bio_unroll=False, |
| image_size=224, |
| hook_type='None', |
| readout_type='multi' |
|
|
| Additionally, the b_vs, vNet, and rn50 trained models are also available (see https://github.com/KietzmannLab/BLT-VS/tree/main/blt_vs_model/training_code/models for model definitions). |
