flaviagiammarino
commited on
Commit
•
a3cb4c5
1
Parent(s):
6a5851a
Update README.md
Browse files
README.md
CHANGED
@@ -15,7 +15,7 @@ This repository is based on the paper, code and pre-trained model released by th
|
|
15 |
MedSAM was trained on a large-scale medical image segmentation dataset of 1,090,486 image-mask pairs collected from different publicly available sources.
|
16 |
The image-mask pairs cover 15 imaging modalities and over 30 cancer types.
|
17 |
|
18 |
-
MedSAM was initialized
|
19 |
The training was performed for 100 epochs with a batch size of 160 using the AdamW optimizer with a learning rate of 10−4 and a weight decay of 0.01.
|
20 |
|
21 |
- **Repository:** [MedSAM Official GitHub Repository](https://github.com/bowang-lab/medsam)
|
|
|
15 |
MedSAM was trained on a large-scale medical image segmentation dataset of 1,090,486 image-mask pairs collected from different publicly available sources.
|
16 |
The image-mask pairs cover 15 imaging modalities and over 30 cancer types.
|
17 |
|
18 |
+
MedSAM was initialized using the pre-trained SAM model with the ViT-Base backbone. The prompt encoder weights were frozen, while the image encoder and mask decoder weights were updated during training.
|
19 |
The training was performed for 100 epochs with a batch size of 160 using the AdamW optimizer with a learning rate of 10−4 and a weight decay of 0.01.
|
20 |
|
21 |
- **Repository:** [MedSAM Official GitHub Repository](https://github.com/bowang-lab/medsam)
|