Semantic Segmentation on ADE20K

Object Detection on COCO

Checkpoints for VMamba (alias of vssm version 0)

These checkpoints correspond to the experiments done before date #20240119.

Most backbone models trained without ema, which do not enhance performance \cite(Swin-Transformer). We use ema because our model is still under development, without hyperparameter tuning.

The checkpoints used in object detection and segmentation is VMamba-B with droppath 0.5 + no ema. VMamba-B* represents for VMamba-B with droppath 0.6 + ema, the performance of which is non-ema: 83.3 in epoch 262; ema: 83.7 in epoch 241

Semantic Segmentation on ADE20K

Object Detection on COCO

• Models in this subsection is initialized from the models trained in classfication.
• we now calculate FLOPs with the algrithm @ albertgu provides, which will be bigger than previous calculation (which is based on the selective_scan_ref function, and ignores the hardware-aware algrithm).

Classification on ImageNet-1K

• Models in this subsection is trained from scratch with random or manual initialization. The hyper-parameters are inherited from Swin, except for drop_path_rate and EMA. All models are trained with EMA except for the Vanilla-VMamba-T.
• TP.(Throughput) and Train TP. (Train Throughput) are assessed on an A100 GPU paired with an AMD EPYC 7542 CPU, with batch size 128. Train TP. is tested with mix-resolution, excluding the time consumption of optimizers.
• FLOPs and parameters are now gathered with head (In previous versions, without head, so the numbers raise a little bit).
• we calculate FLOPs with the algorithm @ albertgu provides, which will be bigger than previous calculation (which is based on the selective_scan_ref function, and ignores the hardware-aware algorithm).

We use ema because our model is still under development, without hyperparameter tuning.

This is a pre-release