Upload upstream_FetReg2021_segmentation_visualisation.py with huggingface_hub

upstream_FetReg2021_segmentation_visualisation.py

@@ -0,0 +1,92 @@
+"""
+Fetoscopy placental vessel segmentation and registration challenge (FetReg)
+EndoVis - MICCAI2021
+Challenge link: https://www.synapse.org/#!Synapse:syn25313156
+Visualization script for image and mask for the semantic segmentation task
+"""
+
+import os
+import cv2
+import numpy as np
+import matplotlib.pyplot as plt
+
+
+def get_colormap():
+    """
+    Returns FetReg colormap
+    """
+    colormap = np.asarray(
+        [
+            [0, 0, 0],   # 0 - background 
+            [255, 0, 0], # 1 - vessel
+            [0, 0, 255], # 2 - tool
+            [0, 255, 0], # 3 - fetus
+
+        ]
+        )
+    return colormap
+
+def plot_image_n_label(img_path_fname, mask_path_fname):
+    """
+    Plot of image and RGB mask for visualisation 
+    Params
+        img_path_fname : Input image path
+        mask_path_fname: Input segmentation mask path
+    Return
+        plot of image and RGB mask   
+    """
+    
+    img = cv2.imread(img_path_fname)
+    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+    mask = cv2.imread(mask_path_fname, cv2.COLOR_BGR2GRAY)
+    
+    colormap = get_colormap()
+    
+    mask_rgb = np.zeros(mask.shape[:2] + (3,), dtype=np.uint8)
+    for cnt in range(len(colormap)):
+        mask_rgb[mask == cnt] = colormap[cnt]
+    
+
+    fig, axs = plt.subplots(1, 2, figsize=(14, 7))
+    axs[0].imshow(img)
+    axs[0].axis("off")
+
+
+    axs[1].imshow(mask_rgb)
+    axs[1].axis("off")
+    fig.tight_layout()
+    #plt.show()
+    
+    return fig
+
+
+if __name__ == "__main__":
+    import argparse
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--root", help="Path to root (video) folder that contains images and labels subfolders", required=True)
+    parser.add_argument("--output", help="Output path to save plot", required=True)
+    args = parser.parse_args()
+
+    assert os.path.isdir(args.root), f"{args.root} directory does not exist"
+    
+    img_path = os.path.join(args.root, 'images')
+    mask_path = os.path.join(args.root, 'labels')
+    assert os.path.exists(img_path), f"{img_path} images/labels do not exist."
+    assert os.path.exists(mask_path), f"{mask_path} images/labels do not exist."
+    
+
+    Img_list = np.sort(os.listdir(img_path)) # List all image names
+    
+
+    for cnt in range(len(Img_list)):
+        fname = Img_list[cnt]
+        img_path_fname = os.path.join(img_path, fname)
+        mask_path_fname = os.path.join(mask_path, fname)
+       
+        fig = plot_image_n_label(img_path_fname, mask_path_fname)
+        if not os.path.isdir(args.output):
+            os.makedirs(args.output)
+        fname2 = fname.replace('png','jpg')
+        fig.savefig(os.path.join(args.output, fname2))
+