created external_sequence for Run30 LV project and updated Readme

README.md

@@ -48,3 +48,11 @@ Ask for help on our mailing list:
 
 
 
+## Working with plugins
+
+Plugins is a system of extensions to PyPTV without the need to change the GUI
+
+1. copy the `external_sequence_list.txt` and `external_tracker_list.txt` to the working folder
+2. copy the `plugins/` directory to the working folder
+3. modify the code so it performs instead of the default sequence or default tracker
+4. Open the GUI and Plugins -> Choose , then run the rest: Init -> Sequence 

bump_version.py

@@ -41,6 +41,8 @@ def increment_version(version, part='minor'):
     elif part == 'minor':
         minor += 1
         patch = 0
+    elif part == 'patch':
+        patch += 1
     else:
         raise ValueError("Invalid part specified. Use 'major' or 'minor'.")
 
@@ -56,7 +58,7 @@ def increment_version(version, part='minor'):
     print(f"Current version is {current_version}")
 
     # Example usage
-    new_version = increment_version(current_version, 'minor')
+    new_version = increment_version(current_version, 'patch')
     print(f"New version is {new_version}")
 
     # Update the version in pyproject.toml

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "pyptv"
-version = "0.3.1"
+version = "0.3.2"
 description = "Python GUI for the OpenPTV library `liboptv`"
 authors = [
     { name = "Alex Liberzon", email = "[email protected]" }

pyptv/__version__.py

@@ -1 +1 @@
-__version__ = "0.3.1"
+__version__ = "0.3.2"

pyptv/external_sequence_list.txt

@@ -1,2 +1,3 @@
-ext_sequence_denis
+plugins/ext_sequence_denis
+plugins/ext_sequence_rembg
 

pyptv/external_tracker_list.txt

@@ -1,2 +1,2 @@
-ext_tracker_denis
+plugins/ext_tracker_denis
 

pyptv/plugins/ext_sequence_rembg.py

@@ -0,0 +1,302 @@
+import random
+
+import numpy as np
+from imageio.v3 import imread, imwrite
+from pathlib import Path
+
+from skimage import img_as_ubyte
+from skimage import filters, measure, morphology
+from skimage.color import rgb2gray, label2rgb
+from skimage.segmentation import clear_border
+from skimage.morphology import binary_erosion, binary_dilation, disk
+from skimage.util import img_as_ubyte
+
+from optv.correspondences import correspondences, MatchedCoords
+from optv.tracker import default_naming
+from optv.orientation import point_positions
+
+import matplotlib.pyplot as plt
+
+
+def mask_image(imname : Path, display: bool = False) -> np.ndarray:
+    """Mask the image using a simple high pass filter.
+    
+    Parameters
+    ----------
+    img : np.ndarray
+        The image to be masked.
+        
+    Returns
+    -------
+    np.ndarray
+        The masked image.
+    """
+
+    img = imread(imname)
+    if img.ndim > 2:
+        img = rgb2gray(img)
+
+    if img.dtype != np.uint8:
+        img = img_as_ubyte(img)
+
+    # Apply Gaussian filter to smooth the image
+    smoothed_frame = filters.gaussian(img, sigma=5)
+
+    if display:
+        plt.figure()
+        plt.imshow(smoothed_frame)
+        plt.show()
+
+    # Apply Otsu's thresholding method to segment the object
+    thresh = filters.threshold_otsu(smoothed_frame)
+    # print('Threshold:', thresh)
+    binary_frame = smoothed_frame > 1.1*thresh
+
+    if display:
+        plt.figure()
+        plt.imshow(binary_frame)
+        plt.show()
+
+
+    # binary_frame_cleared = clear_border(binary_frame, buffer_size=20)
+    binary_frame_cleared = binary_frame.copy()
+
+    # plt.figure()
+    # plt.imshow(binary_frame_cleared)
+    # plt.show()
+
+    # Remove small bright objects
+    cleaned_frame = morphology.remove_small_objects(binary_frame_cleared, min_size=100000)
+
+    # %%
+    # Apply morphological closing to close the boundary
+    closed_cleaned_frame = binary_dilation(cleaned_frame, disk(21))
+    closed_cleaned_frame = binary_erosion(closed_cleaned_frame, disk(21))
+
+    if display:
+        # Display the result
+        plt.figure()
+        plt.imshow(closed_cleaned_frame, cmap='gray')
+        plt.title('Closed Boundary of Cleaned Frame')
+        plt.show()
+
+
+    # check the size of the second largest black hole
+    # labeled_frame = measure.label(~closed_cleaned_frame)
+    # regions = measure.regionprops(labeled_frame)
+    # areas = np.array([r.area for r in regions])
+    # area_to_remove = np.sort(areas)[-2] # 2nd largest, 1st is the surrounding
+
+    # %%
+    # Fill holes inside the binary frame to remove large black objects
+    filled_frame = morphology.remove_small_holes(closed_cleaned_frame, area_threshold=2e6)
+
+    if display:
+        # # Display the result
+        plt.figure()
+        plt.imshow(filled_frame, cmap='gray')
+        plt.title('Binary Frame with Large Black Objects Removed')
+        plt.show()
+
+    # %%
+
+    # # Remove small objects and clear the border
+    # cleaned_frame = morphology.remove_small_objects(binary_frame, min_size=100000)
+    # # Fill holes inside the binary frame to remove dark islands
+    # filled_frame = morphology.remove_small_holes(cleaned_frame, area_threshold=100000)
+
+    # filled_frame = clear_border(filled_frame)
+
+    # Label the segmented regions
+    labeled_frame = measure.label(filled_frame)
+
+    if display:
+        # Show the labeled filled frame as a color labeled image
+        plt.figure()
+        plt.imshow(label2rgb(labeled_frame, image=img, bg_label=0))
+        plt.title('Color Labeled Frame with Filled Holes')
+        plt.show()
+
+    # %%
+
+    # Find region properties
+    regions = measure.regionprops(labeled_frame)
+
+    # Assuming the largest region is the object of interest
+    largest_region = max(regions, key=lambda r: r.area)
+
+
+    # Find the smooth contour that surrounds the largest region
+    smooth_contour = morphology.convex_hull_image(largest_region.image)
+
+    # Create an empty image to draw the smooth contour
+    smooth_contour_image = np.zeros_like(labeled_frame, dtype=bool)
+
+    # Place the smooth contour in the correct location
+    minr, minc, maxr, maxc = largest_region.bbox
+    smooth_contour_image[minr:maxr, minc:maxc] = smooth_contour
+
+    if display:
+        # Display the smooth contour on the labeled image
+        plt.figure()
+        plt.imshow(labeled_frame, cmap='jet')
+        plt.contour(smooth_contour_image, colors='red', linewidths=2)
+        plt.title(f'Segmented Object with Smooth Contour')
+        plt.show()
+
+
+    # Convert the largest region to a black and white image
+    bw_image = np.zeros_like(labeled_frame, dtype=bool)
+    bw_image[largest_region.coords[:, 0], largest_region.coords[:, 1]] = True
+
+    # plt.figure(), plt.imshow(bw_image, cmap='gray')
+
+    # Apply morphological closing to remove sharp spikes
+    closed_image = binary_dilation(bw_image, disk(21))
+    closed_image = binary_erosion(closed_image, disk(21))
+
+    if display:
+        # Display the result
+        plt.figure()
+        plt.imshow(closed_image, cmap='gray')
+        plt.title('Smooth Boundary without Sharp Spikes')
+        plt.show()
+
+
+    # Apply morphological operations to get the external contour
+    eroded_image = binary_erosion(closed_image, disk(1))
+    external_contour = closed_image & ~eroded_image
+
+    imwrite(imname.with_suffix('.jpg'), img_as_ubyte(external_contour))
+
+    # Dilate the external contour for better visibility
+    dilated_external_contour = binary_dilation(external_contour, disk(3))
+
+    # Create a masked image of the same size as the input image
+    masked_image = np.zeros_like(img, dtype=np.uint8)
+    # Mask out (black) everything outside of closed_image
+    masked_image[closed_image] = img[closed_image]
+
+    if display:
+        plt.figure()
+        plt.imshow(masked_image)
+        plt.show()
+
+    return masked_image
+
+class Sequence:
+    """Sequence class defines external tracking addon for pyptv
+    User needs to implement the following functions:
+            do_sequence(self)
+
+    Connection to C ptv module is given via self.ptv and provided by pyptv software
+    Connection to active parameters is given via self.exp1 and provided by pyptv software.
+
+    User responsibility is to read necessary files, make the calculations and write the files back.
+    """
+
+    def __init__(self, ptv=None, exp=None):
+        self.ptv = ptv
+        self.exp = exp
+
+    def do_sequence(self):
+        """ Copy of the sequence loop with one change we call everything as 
+        self.ptv instead of ptv. 
+        
+        """
+        # Sequence parameters    
+
+        n_cams, cpar, spar, vpar, tpar, cals = (
+            self.exp.n_cams,
+            self.exp.cpar,
+            self.exp.spar,
+            self.exp.vpar,
+            self.exp.tpar,
+            self.exp.cals,
+        )
+
+        # # Sequence parameters
+        # spar = SequenceParams(num_cams=n_cams)
+        # spar.read_sequence_par(b"parameters/sequence.par", n_cams)
+
+
+        # sequence loop for all frames
+        first_frame = spar.get_first()
+        last_frame = spar.get_last()
+        print(f" From {first_frame = } to {last_frame = }")
+
+        for frame in range(first_frame, last_frame + 1):
+            # print(f"processing {frame = }")
+
+            detections = []
+            corrected = []
+            for i_cam in range(n_cams):
+                base_image_name = spar.get_img_base_name(i_cam).decode()
+                imname = Path(base_image_name % frame) # works with jumps from 1 to 10 
+                masked_image = mask_image(imname)
+
+                # img = imread(imname)
+                # if img.ndim > 2:
+                #     img = rgb2gray(img)
+
+                # if img.dtype != np.uint8:
+                #     img = img_as_ubyte(img)
+
+
+
+                high_pass = self.ptv.simple_highpass(masked_image, cpar)
+                targs = self.ptv.target_recognition(high_pass, tpar, i_cam, cpar)
+
+                targs.sort_y()
+                detections.append(targs)
+                masked_coords = MatchedCoords(targs, cpar, cals[i_cam])
+                pos, _ = masked_coords.as_arrays()
+                corrected.append(masked_coords)
+
+            #        if any([len(det) == 0 for det in detections]):
+            #            return False
+
+            # Corresp. + positions.
+            sorted_pos, sorted_corresp, _ = correspondences(
+                detections, corrected, cals, vpar, cpar)
+
+            # Save targets only after they've been modified:
+            # this is a workaround of the proper way to construct _targets name
+            for i_cam in range(n_cams):
+                base_name = spar.get_img_base_name(i_cam).decode()
+                # base_name = replace_format_specifiers(base_name) # %d to %04d
+                self.ptv.write_targets(detections[i_cam], base_name, frame)
+
+            print("Frame " + str(frame) + " had " +
+                repr([s.shape[1] for s in sorted_pos]) + " correspondences.")
+
+            # Distinction between quad/trip irrelevant here.
+            sorted_pos = np.concatenate(sorted_pos, axis=1)
+            sorted_corresp = np.concatenate(sorted_corresp, axis=1)
+
+            flat = np.array([
+                corrected[i].get_by_pnrs(sorted_corresp[i])
+                for i in range(len(cals))
+            ])
+            pos, _ = point_positions(flat.transpose(1, 0, 2), cpar, cals, vpar)
+
+            # if len(cals) == 1: # single camera case
+            #     sorted_corresp = np.tile(sorted_corresp,(4,1))
+            #     sorted_corresp[1:,:] = -1
+
+            if len(cals) < 4:
+                print_corresp = -1 * np.ones((4, sorted_corresp.shape[1]))
+                print_corresp[:len(cals), :] = sorted_corresp
+            else:
+                print_corresp = sorted_corresp
+
+            # Save rt_is
+            rt_is_filename = default_naming["corres"].decode()
+            rt_is_filename = rt_is_filename + f'.{frame}'
+            with open(rt_is_filename, "w", encoding="utf8") as rt_is:
+                rt_is.write(str(pos.shape[0]) + "\n")
+                for pix, pt in enumerate(pos):
+                    pt_args = (pix + 1, ) + tuple(pt) + tuple(print_corresp[:, pix])
+                    rt_is.write("%4d %9.3f %9.3f %9.3f %4d %4d %4d %4d\n" % pt_args)
+
+