Table 3.
Processing Time and Model Parameter Statistics for SAM and SAM2 Models.
| mobile_SAM | SAM_b | SAM_l | SAM2_t | SAM2_s | SAM2_b | SAM2_l | |
|---|---|---|---|---|---|---|---|
| Parameters | 10.1 (M) | 93.7 (M) | 312.3 (M) | 38.9 (M) | 46.0 (M) | 80.8 (M) | 224.4 (M) |
| Total time* | 1876 (S) | 6482 (S) | 14643 (S) | 3142 (S) | 3434 (S) | 5080 (S) | 9987 (S) |
* The total time includes the combined inference time for the image encoder, prompt encoder, and mask decoder when processing 2,000 images from the ADE20K dataset. Inference was performed using a single NVIDIA RTX 3090 GPU.