Visualize show instances

docs/visualize_display_instances.md

@@ -0,0 +1,64 @@
+## Display Instances
+
+This function displays different object intances in different colors on top of the original image.
+
+**plantcv.visualize.display_instances**(*img, masks, figsize=(16, 16), title="", colors=None, captions=None, show_bbox=True, ax=None*)
+**returns** masked_img, colors
+
+- **Parameters:**
+ - img - (required, ndarray)input image
+ - masks - (required, ndarray) instance masks represented by a 3-d array, the 3rd dimension represents the number of insatnces to show.
+ - figsize - (optional, tuple) the size of the generated figure
+ - title - (optional, str) the title of the figure
+ - colors - (optional, list of tuples, every value should be in the range of [0.0,1.0]) a list of colors to use with each object. If no value is passed, a set of random colors would be used
+ - captions - (optional, str) a list of strings to use as captions for each object. If no list of captions is provided, show the local index of the instance
+ - show_bbox - (optional, bool) indicator of whether showing the bounding-box
+ - ax - (optional, matplotlib.axes._subplots.AxesSubplot) the axis to plot on. If no axis is passed, a new one will be created 
+- **Context:**
+ - Used to display different segmented instances on top of the original image. 
+- **Example use:**
+ - Below
+
+**Original image: RGB image**
+
+![Screenshot](img/documentation_images/visualize_display_instances/visualize_inst_seg_img.png)
+
+**masks: 10 segmentation masks represent for different leaves**
+
+![Screenshot](img/documentation_images/visualize_display_instances/mask_0.png)
+![Screenshot](img/documentation_images/visualize_display_instances/mask_1.png)
+![Screenshot](img/documentation_images/visualize_display_instances/mask_2.png)
+![Screenshot](img/documentation_images/visualize_display_instances/mask_3.png)
+![Screenshot](img/documentation_images/visualize_display_instances/mask_4.png)
+![Screenshot](img/documentation_images/visualize_display_instances/mask_5.png)
+![Screenshot](img/documentation_images/visualize_display_instances/mask_6.png)
+![Screenshot](img/documentation_images/visualize_display_instances/mask_7.png)
+![Screenshot](img/documentation_images/visualize_display_instances/mask_8.png)
+![Screenshot](img/documentation_images/visualize_display_instances/mask_9.png)
+![Screenshot](img/documentation_images/visualize_display_instances/mask_10.png)
+
+
+```python
+
+from plantcv import plantcv as pcv
+
+masked_img,colors = pcv.visualize.display_instances(img, masks, figsize=(10, 10), title="", ax=None, colors=None, captions=None, show_bbox=True)
+
+# option to add customized captions and/or customized figure title and/or not showing the bounding box
+captions = ["leaf{}".format(i) for i in range(masks.shape[2])]
+_,_ = pcv.visualize.display_instances(img, masks, figsize=(16, 16), title="Visualization of segmentation", ax=None, colors=None, captions=captions, show_bbox=False)
+
+# the number of instances to display depends on the number of input masks
+# the colors can also be customized 
+masks_reduced = masks[:,:,2:7],
+colors = [(0.0,1.0,0.2),(0.4,0.0,1.0),(0.5,1.0,0.0),(1.0,0.6,0.0),(0.9,0.0,1.0)]
+_,_= pcv.visualize.display_instances(img, masks_reduced, figsize=(16, 16), title="", ax=None, colors=colors, captions=None, show_bbox=True)
+
+```
+
+**Blended Image**
+
+![Screenshot](img/documentation_images/visualize_display_instances/result1.png)
+![Screenshot](img/documentation_images/visualize_display_instances/result2.png)
+![Screenshot](img/documentation_images/visualize_display_instances/result3.png)
+**Source Code:** [Here](https://github.com/danforthcenter/plantcv/blob/master/plantcv/plantcv/visualize/display_instances.py)

plantcv/plantcv/hyperspectral/read_data.py

@@ -118,16 +118,19 @@ def read_data(filename):
  hdata = hdata.replace("{\n", "{")
  hdata = hdata.replace("\n}", "}")
  hdata = hdata.replace(" \n ", "")
+ hdata = hdata.replace(" \n", "")
  hdata = hdata.replace(";", "")
  hdata = hdata.split("\n")
 
  # Loop through and create a dictionary from the header file
  for i, string in enumerate(hdata):
  if ' = ' in string:
  header_data = string.split(" = ")
+ header_data[0] = header_data[0].lower()
  header_dict.update({header_data[0].rstrip(): header_data[1].rstrip()})
  elif ' : ' in string:
  header_data = string.split(" : ")
+ header_data[0] = header_data[0].lower()
  header_dict.update({header_data[0].rstrip(): header_data[1].rstrip()})
 
  # Reformat wavelengths

plantcv/plantcv/visualize/__init__.py

@@ -5,10 +5,12 @@
 from plantcv.plantcv.visualize.colorspaces import colorspaces
 from plantcv.plantcv.visualize.auto_threshold_methods import auto_threshold_methods
 from plantcv.plantcv.visualize.overlay_two_imgs import overlay_two_imgs
+from plantcv.plantcv.visualize.display_instances import display_instances
 from plantcv.plantcv.visualize.colorize_label_img import colorize_label_img
 from plantcv.plantcv.visualize.obj_sizes import obj_sizes
 from plantcv.plantcv.visualize.obj_size_ecdf import obj_size_ecdf
 from plantcv.plantcv.visualize.hyper_histogram import hyper_histogram
 
 __all__ = ["pseudocolor", "colorize_masks", "histogram", "clustered_contours", "colorspaces", "auto_threshold_methods",
- "overlay_two_imgs", "colorize_label_img", "obj_size_ecdf", "obj_sizes", "hyper_histogram"]
+ "overlay_two_imgs","display_instances", "colorize_label_img", "obj_size_ecdf", "obj_sizes", "hyper_histogram"]
+

plantcv/plantcv/visualize/display_instances.py

@@ -0,0 +1,152 @@
+# display instances in an image
+import cv2
+from matplotlib.patches import Polygon
+import colorsys
+import random
+from skimage.measure import find_contours
+import matplotlib.pyplot as plt
+import numpy as np
+from matplotlib import patches, lines
+from matplotlib.patches import Polygon
+from plantcv.plantcv import fatal_error, params, color_palette
+import os
+from cv2 import cvtColor, COLOR_BGR2RGB
+
+def _overlay_mask_on_img(img, mask, color, alpha=0.5):
+ """ Overlay a given mask on top of an image such that the masked area (the non-zero areas in the mask) is shown in user
+ defined color, the other area (the zero-valued areas in the mask) is shown in original image.
+ Inputs:
+ img = image to show, can be either visible or grayscale
+ mask = mask to be put on top of the image
+ color = a tuple of desired color to show the mask on top of the image
+ alpha = a value (between 0 and 1) indicating the transparency when blending mask and image, by default 0.5
+ Output:
+ img = the original image with mask overlaied on top
+
+ :param img: numpy.ndarray
+ :param mask: numpy.ndarray
+ :param color: tuple
+ :param alpha: float
+ :return: img: numpy.ndarray
+ """
+ # if len(img.shape) == 2:
+ # img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
+ for c in range(img.shape[-1]):
+ img[:, :, c] = np.where(mask == 1,
+ img[:, :, c] *
+ (1 - alpha) + alpha * color[c] * 255,
+ img[:, :, c])
+ return img
+
+
+def display_instances(img, masks, figsize=(16, 16), title="", colors=None, captions=None, show_bbox=True, ax=None):
+ """ Display multiple instances in image based on the given masks
+ This function is inspired by the same function used in mrcnn, showing different instances with different colors on top of the original image
+ Users also have the option to specify the color for every instance, as well as the caption for every instance. Showing bounding boxes is another option.
+ Inputs:
+ img = rgb image to show
+ mask = a bunch of masks in a matrix
+ figsize = desired size of figure, by default (16,16)
+ title = desired title to show, by default ""
+ colors = a list of colors for every instance to display, by default colors=None
+ captions = a list of names for every instance, by default captions=None
+ show_bbox = a flog indicating whether show bounding box, by default show_bbox=True
+ ax = axis to show, by default ax=None
+ Outputs:
+ masked_img = image with instances masks overlaied on top
+ colors = colors used to display every instance
+
+ :param img: numpy.ndarray
+ :param masks: numpy.ndarray
+ :param figsize: tuple
+ :param title: str
+ :param colors: list (of tuples)
+ :param captions: str
+ :param show_bbox: bool
+ :param ax:
+ :return: masked_img: numpy.ndarray
+ :return: colors: list (of tuples)
+ """
+
+ debug = params.debug
+ params.debug = None
+
+ # Auto-increment the device counter
+ params.device += 1
+
+ if img.shape[0:2] != masks.shape[0:2]:
+ fatal_error("Sizes of image and mask mismatch!")
+ #
+ # auto_show = False
+ if not ax:
+ _, ax = plt.subplots(1, figsize=figsize)
+ # auto_show = True
+
+ num_insts = masks.shape[2]
+
+ # # Generate random colors
+ # colors = colors or _random_colors(num_insts)
+
+ if colors is not None:
+ if max(max(colors)) > 1 or min(min(colors)) < 0:
+ fatal_error("RGBA values should be within 0-1 range!")
+ else:
+ colors_ = color_palette(num_insts)
+ colors = [tuple([ci / 255 for ci in c]) for c in colors_]
+
+ if len(colors) < num_insts:
+ fatal_error("Not enough colors provided to show all instances!")
+ if len(colors) > num_insts:
+ colors = colors[0:num_insts]
+
+ # Show area outside image boundaries.
+ height, width = img.shape[:2]
+ ax.set_ylim(height + 10, -10)
+ ax.set_xlim(-10, width + 10)
+ ax.axis('off')
+ ax.set_title(title)
+
+ masked_img = img.astype(np.uint32).copy()
+ for i in range(num_insts):
+ color = colors[i]
+
+ # Mask
+ mask = masks[:, :, i]
+ # the color is in order of r-g-b, however the image is in order of b-g-r, so take the reverse of color
+ masked_img = _overlay_mask_on_img(masked_img, mask, color[::-1])
+
+ ys, xs = np.where(mask > 0)
+ x1, x2 = min(xs), max(xs)
+ y1, y2 = min(ys), max(ys)
+ if show_bbox:
+ p = patches.Rectangle((x1, y1), x2 - x1, y2 - y1, linewidth=2, alpha=0.7, linestyle="dashed", edgecolor=color, facecolor='none')
+ ax.add_patch(p)
+
+ # Mask Polygon
+ # Pad to ensure proper polygons for masks that touch image edges.
+ padded_mask = np.zeros((mask.shape[0] + 2, mask.shape[1] + 2), dtype=np.uint8)
+ padded_mask[1:-1, 1:-1] = mask
+ contours = find_contours(padded_mask, 0.5)
+ for verts in contours:
+ # Subtract the padding and flip (y, x) to (x, y)
+ verts = np.fliplr(verts) - 1
+ p = Polygon(verts, facecolor="none", edgecolor=color)
+ ax.add_patch(p)
+ if not captions:
+ caption = str(i)
+ else:
+ caption = captions[i]
+ ax.text(x1, y1 + 8, caption, color='w', size=13, backgroundcolor="none")
+
+ masked_img = cvtColor(masked_img.astype(np.uint8), COLOR_BGR2RGB)
+ params.debug = debug
+ if params.debug is not None:
+ if params.debug == "plot":
+ ax.imshow(masked_img)
+
+ if params.debug == "print":
+ ax.imshow(masked_img)
+ plt.savefig(os.path.join(params.debug_outdir, str(params.device) + "_displayed_instances.png"))
+ plt.close("all")
+
+ return masked_img, colors

tests/tests.py

@@ -20,8 +20,10 @@
 import matplotlib.pyplot as plt
 import dask
 from dask.distributed import Client
+import pickle as pkl
 from skimage import img_as_ubyte
 
+
 PARALLEL_TEST_DATA = os.path.join(os.path.dirname(os.path.abspath(__file__)), "parallel_data")
 TEST_TMPDIR = os.path.join(os.path.dirname(os.path.abspath(__file__)), "..", ".cache")
 TEST_IMG_DIR = "images"
@@ -1082,6 +1084,8 @@ def test_plantcv_parallel_process_results_invalid_json():
 # TEST_IM_BAD_NAN = "bad_mask_nan.pkl"
 # TEST_IM_BAD_INF = "bad_mask_inf.pkl"
 PIXEL_VALUES = "pixel_inspector_rgb_values.txt"
+TEST_INPUT_INSTANCE_IMG = "visualize_inst_seg_img.png"
+TEST_INPUT_INSTANCE_MASK = "visualize_inst_seg_mask.pkl"
 TEST_INPUT_LEAF_MASK = "leaves_mask.png"
 
 # leaving photosynthesis data here so it can be used to test plot_image and print_image
@@ -6766,6 +6770,64 @@ def test_plantcv_visualize_overlay_two_imgs_size_mismatch():
  with pytest.raises(RuntimeError):
  _ = pcv.visualize.overlay_two_imgs(img1=img1, img2=img2)
 
+def test_plantcv_visualize_display_instances():
+ cache_dir = os.path.join(TEST_TMPDIR, "test_plantcv_visualize_display_instances")
+ os.mkdir(cache_dir)
+ # create synthetic test data
+ img = img_as_ubyte(np.random.rand(10,10,3))
+ masks = np.zeros((10, 10, 2))
+ masks[1:3, 3:5, 0] = 1
+ masks[2:5, 1:3, 1] = 1
+ colors = [(1.0,0.0,0.15),(1.0,0.0,0.74)]
+ pcv.params.debug = "plot"
+ _, ax = plt.subplots(1, 1, figsize=(16, 16))
+ _, colors = pcv.visualize.display_instances(img, masks, title="test", colors=colors, captions=['1','2'], show_bbox=True, ax=ax)
+ pcv.params.debug = "print"
+ pcv.params.debug_outdir = cache_dir
+ _, colors = pcv.visualize.display_instances(img, masks)
+ assert len(colors) == masks.shape[2]
+
+def test_plantcv_visualize_display_instances_bad_color():
+ pcv.params.debug = None
+ # create synthetic test data
+ img = img_as_ubyte(np.random.rand(10,10,3))
+ masks = np.zeros((10, 10, 2), dtype=np.uint8)
+ masks[1:3, 3:5, 0] = 1
+ masks[2:5, 1:3, 1] = 1
+ colors = [(1.0,0.0,0.15)]
+ with pytest.raises(RuntimeError):
+ _, _ = pcv.visualize.display_instances(img, masks, colors=colors)
+
+def test_plantcv_visualize_display_instances_bad_color2():
+ pcv.params.debug = None
+ # create synthetic test data
+ img = img_as_ubyte(np.random.rand(10,10,3))
+ masks = np.zeros((10, 10, 2), dtype=np.uint8)
+ masks[1:3, 3:5, 0] = 1
+ masks[2:5, 1:3, 1] = 1
+ colors = [(1.0,0.0,0.15),(1.0,0.0,0.74),(0.0,1.0,0.74)]
+ _, colors = pcv.visualize.display_instances(img, masks, colors=colors)
+ assert len(colors) == masks.shape[2]
+
+def test_plantcv_visualize_display_instances_bad_color3():
+ pcv.params.debug = None
+ # create synthetic test data
+ img = img_as_ubyte(np.random.rand(10,10,3))
+ masks = np.zeros((10, 10, 2), dtype=np.uint8)
+ masks[1:3, 3:5, 0] = 1
+ masks[2:5, 1:3, 1] = 1
+ colors = [(10,0.0,15.0),(10,0.0,74)]
+ with pytest.raises(RuntimeError):
+ _, _ = pcv.visualize.display_instances(img, masks, colors=colors)
+
+def test_plantcv_visualize_display_instances_bad_size():
+ pcv.params.debug = None
+ # create synthetic test data
+ img = img_as_ubyte(np.random.rand(5,5,3))
+ masks = np.zeros((10, 10, 2), dtype=np.uint8)
+ with pytest.raises(RuntimeError):
+ _, _ = pcv.visualize.display_instances(img, masks)
+
 
 @pytest.mark.parametrize("num,expected", [[100, 35], [30, 33]])
 def test_plantcv_visualize_size(num, expected):