Upload folder using huggingface_hub
Browse files- MedSAM2_pretrain_10ep_b1_AMD-SD_sam2_hiera_t.pth +3 -0
- README.md +0 -3
- config.json +5 -0
- inference.py +39 -0
- requirements.txt +6 -0
MedSAM2_pretrain_10ep_b1_AMD-SD_sam2_hiera_t.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8018821b159767012d081db1f5cef2720aa495bad6861b8d6b1c63dce24a5a6e
|
3 |
+
size 155972738
|
README.md
CHANGED
@@ -1,3 +0,0 @@
|
|
1 |
-
---
|
2 |
-
license: apache-2.0
|
3 |
-
---
|
|
|
|
|
|
|
|
config.json
ADDED
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"pipeline_tag": "image-segmentation",
|
3 |
+
"base_model": "medsam2"
|
4 |
+
}
|
5 |
+
|
inference.py
ADDED
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from typing import Dict
|
2 |
+
import torch
|
3 |
+
import numpy as np
|
4 |
+
from PIL import Image
|
5 |
+
from skimage import transform
|
6 |
+
from sam2.build_sam import build_sam2
|
7 |
+
from sam2.sam2_image_predictor import SAM2ImagePredictor
|
8 |
+
|
9 |
+
class PreTrainedModel:
|
10 |
+
def __init__(self):
|
11 |
+
self.model = build_sam2(
|
12 |
+
"sam2_hiera_t",
|
13 |
+
"MedSAM2_pretrain_10ep_b1_AMD-SD_sam2_hiera_t.pth",
|
14 |
+
device="cuda" if torch.cuda.is_available() else "cpu"
|
15 |
+
)
|
16 |
+
self.predictor = SAM2ImagePredictor(self.model)
|
17 |
+
|
18 |
+
def __call__(self, inputs: Dict):
|
19 |
+
image = Image.open(inputs["image"]).convert("RGB")
|
20 |
+
box = list(map(float, inputs["box"]))
|
21 |
+
|
22 |
+
image_np = np.array(image)
|
23 |
+
img_3c = image_np if image_np.shape[2] == 3 else np.repeat(image_np[:, :, None], 3, axis=-1)
|
24 |
+
img_1024 = transform.resize(img_3c, (1024, 1024), preserve_range=True).astype(np.uint8)
|
25 |
+
|
26 |
+
box_1024 = np.array(box) / [image_np.shape[1], image_np.shape[0], image_np.shape[1], image_np.shape[0]] * 1024
|
27 |
+
box_1024 = box_1024[None, :]
|
28 |
+
|
29 |
+
with torch.inference_mode(), torch.autocast("cuda" if torch.cuda.is_available() else "cpu", dtype=torch.bfloat16):
|
30 |
+
self.predictor.set_image(img_1024)
|
31 |
+
masks, _, _ = self.predictor.predict(
|
32 |
+
point_coords=None,
|
33 |
+
point_labels=None,
|
34 |
+
box=box_1024,
|
35 |
+
multimask_output=False
|
36 |
+
)
|
37 |
+
|
38 |
+
mask = masks[0].astype(np.uint8)
|
39 |
+
return {"mask": mask.tolist()}
|
requirements.txt
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
torch
|
2 |
+
numpy
|
3 |
+
scikit-image
|
4 |
+
gradio
|
5 |
+
pillow
|
6 |
+
git+https://github.com/facebookresearch/sam2.git
|