TMAx 0.1.0__tar.gz

tmax-0.1.0/PKG-INFO

@@ -0,0 +1,28 @@
+Metadata-Version: 2.4
+Name: TMAx
+Version: 0.1.0
+Summary: Module for segmenting pathological images of tissue microarrays
+Author-email: BOSCH Valentin <val.bosch2001@gmail.com>
+Requires-Python: >=3.8
+Description-Content-Type: text/markdown
+Requires-Dist: torch
+Requires-Dist: numpy
+Requires-Dist: huggingface_hub
+Requires-Dist: opencv-python
+Requires-Dist: pillow
+
+<img width="2242" height="1267" alt="image" src="https://github.com/user-attachments/assets/3e0d6e0b-aafa-400b-a129-63b837c2ee01" />
+
+# TMAs 
+Ce projet propose un modèle de segmentation d’images histopathologiques, spécialisé dans les Tissue MicroArrays (TMA) utilisés en anatomopathologie.
+
+L’objectif est de segmenter automatiquement des structures d’intérêt dans des lames numériques, en particulier dans le contexte des lymphomes.
+
+# Documentation
+def predict_mask(model, imagePath, outputPath, repo_id="Vaaaal/TMAs", filename="tmas_segmenter.pth"):
+- model (torch.nn.Module)
+- Segmentation model (UNet architecture or equivalent) : Weights will be automatically loaded from Hugging Face.
+- imagePath (str) : Path to the input image to be segmented.
+- outputPath (str) : Path where the generated segmentation mask will be saved.
+- repo_id (str, optional) : ID of the Hugging Face repository containing the model weights.
+- filename (str, optional) : Name of the weight file in the repository.

tmax-0.1.0/README.md

@@ -0,0 +1,15 @@
+<img width="2242" height="1267" alt="image" src="https://github.com/user-attachments/assets/3e0d6e0b-aafa-400b-a129-63b837c2ee01" />
+
+# TMAs 
+Ce projet propose un modèle de segmentation d’images histopathologiques, spécialisé dans les Tissue MicroArrays (TMA) utilisés en anatomopathologie.
+
+L’objectif est de segmenter automatiquement des structures d’intérêt dans des lames numériques, en particulier dans le contexte des lymphomes.
+
+# Documentation
+def predict_mask(model, imagePath, outputPath, repo_id="Vaaaal/TMAs", filename="tmas_segmenter.pth"):
+- model (torch.nn.Module)
+- Segmentation model (UNet architecture or equivalent) : Weights will be automatically loaded from Hugging Face.
+- imagePath (str) : Path to the input image to be segmented.
+- outputPath (str) : Path where the generated segmentation mask will be saved.
+- repo_id (str, optional) : ID of the Hugging Face repository containing the model weights.
+- filename (str, optional) : Name of the weight file in the repository.

tmax-0.1.0/TMAx/__init__.py

@@ -0,0 +1 @@
+from .core import predict_mask

tmax-0.1.0/TMAx/core.py

@@ -0,0 +1,84 @@
+import torch
+import cv2
+import numpy as np
+from huggingface_hub import hf_hub_download, login
+from PIL import Image
+import config
+from model_resnet import UNetResNet
+
+
+#Main variables
+DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
+INPUT_IMAGE_WIDTH = 1024
+INPUT_IMAGE_HEIGHT = 1024
+
+
+def load_model_from_hf(repo_id, filename, device):
+    model = UNetResNet()
+
+    path = hf_hub_download(
+        repo_id=repo_id,
+        filename=filename,
+        repo_type="model"
+    )
+
+    state_dict = torch.load(path, map_location=device)
+    model.load_state_dict(state_dict, strict=False)
+
+    model.to(device)
+    model.eval()
+
+    return model
+
+def predict_mask(imagePath, output_path):
+
+    model = load_model_from_hf(
+    repo_id="Vaaaal/TMAs",
+    filename="unet_tgs_salt_resnet.pth",
+    device=config.DEVICE
+    )
+
+    with torch.no_grad():
+
+        # --- Load image (robuste TIFF) ---
+        try:
+            image = np.array(Image.open(imagePath).convert("RGB"))
+        except Exception as e:
+            raise FileNotFoundError(f"Erreur chargement image: {imagePath}") from e
+
+        origH, origW = image.shape[:2]
+
+        # --- Preprocess ---
+        image_resized = cv2.resize(
+            image,
+            (config.INPUT_IMAGE_WIDTH, config.INPUT_IMAGE_HEIGHT)
+        )
+
+        image_resized = image_resized.astype("float32") / 255.0
+        image_resized = np.transpose(image_resized, (2, 0, 1))
+        image_resized = np.expand_dims(image_resized, 0)
+
+        image_tensor = torch.from_numpy(image_resized).to(config.DEVICE)
+
+        # --- Prediction ---
+        predMask = model(image_tensor)
+
+        # --- Post-process (important) ---
+        predMask = torch.sigmoid(predMask)  # si binaire
+        predMask = predMask.squeeze().cpu().numpy()
+
+        # resize back original size
+        predMask = cv2.resize(predMask, (origW, origH))
+
+        # threshold (optionnel)
+        predMask = (predMask > 0.5).astype("uint8") * 255
+
+        Image.fromarray(predMask).save(output_path, compression=None)
+
+    return predMask
+
+
+# --- IMAGE ---
+img_path = r"D:\imvia\unet_tma_segmenter\TMAs\TMAs\09abe43a32f34fbb98d6ab10ff13ca16.tiff"
+# --- PREDICT ---
+mask = predict_mask(img_path, "predicted_mask.tiff")

tmax-0.1.0/TMAx.egg-info/PKG-INFO

@@ -0,0 +1,28 @@
+Metadata-Version: 2.4
+Name: TMAx
+Version: 0.1.0
+Summary: Module for segmenting pathological images of tissue microarrays
+Author-email: BOSCH Valentin <val.bosch2001@gmail.com>
+Requires-Python: >=3.8
+Description-Content-Type: text/markdown
+Requires-Dist: torch
+Requires-Dist: numpy
+Requires-Dist: huggingface_hub
+Requires-Dist: opencv-python
+Requires-Dist: pillow
+
+<img width="2242" height="1267" alt="image" src="https://github.com/user-attachments/assets/3e0d6e0b-aafa-400b-a129-63b837c2ee01" />
+
+# TMAs 
+Ce projet propose un modèle de segmentation d’images histopathologiques, spécialisé dans les Tissue MicroArrays (TMA) utilisés en anatomopathologie.
+
+L’objectif est de segmenter automatiquement des structures d’intérêt dans des lames numériques, en particulier dans le contexte des lymphomes.
+
+# Documentation
+def predict_mask(model, imagePath, outputPath, repo_id="Vaaaal/TMAs", filename="tmas_segmenter.pth"):
+- model (torch.nn.Module)
+- Segmentation model (UNet architecture or equivalent) : Weights will be automatically loaded from Hugging Face.
+- imagePath (str) : Path to the input image to be segmented.
+- outputPath (str) : Path where the generated segmentation mask will be saved.
+- repo_id (str, optional) : ID of the Hugging Face repository containing the model weights.
+- filename (str, optional) : Name of the weight file in the repository.

tmax-0.1.0/TMAx.egg-info/SOURCES.txt

@@ -0,0 +1,9 @@
+README.md
+pyproject.toml
+TMAx/__init__.py
+TMAx/core.py
+TMAx.egg-info/PKG-INFO
+TMAx.egg-info/SOURCES.txt
+TMAx.egg-info/dependency_links.txt
+TMAx.egg-info/requires.txt
+TMAx.egg-info/top_level.txt

tmax-0.1.0/TMAx.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

tmax-0.1.0/TMAx.egg-info/requires.txt

@@ -0,0 +1,5 @@
+torch
+numpy
+huggingface_hub
+opencv-python
+pillow

tmax-0.1.0/TMAx.egg-info/top_level.txt

@@ -0,0 +1 @@
+TMAx

tmax-0.1.0/pyproject.toml

@@ -0,0 +1,20 @@
+[project]
+name = "TMAx"
+version = "0.1.0"
+description = "Module for segmenting pathological images of tissue microarrays"
+dependencies = [
+    "torch",
+    "numpy",
+    "huggingface_hub",
+    "opencv-python",
+    "pillow"
+]
+authors = [
+    { name="BOSCH Valentin", email="val.bosch2001@gmail.com" }
+]
+readme = "README.md"
+requires-python = ">=3.8"
+
+[build-system]
+requires = ["setuptools", "wheel"]
+build-backend = "setuptools.build_meta"

tmax-0.1.0/setup.cfg

@@ -0,0 +1,4 @@
+[egg_info]
+tag_build = 
+tag_date = 0
+