Update `heatmaps` solution (#16720)
Co-authored-by: UltralyticsAssistant <web@ultralytics.com>pull/16722/head^2
Muhammad Rizwan Munawar
2 months ago
committed by
GitHub
parent
e56f9140fb
commit
4884311991
6 changed files with 142 additions and 353 deletions
@ -1,249 +1,93 @@ |
||||
# Ultralytics YOLO 🚀, AGPL-3.0 license |
||||
|
||||
from collections import defaultdict |
||||
|
||||
import cv2 |
||||
import numpy as np |
||||
|
||||
from ultralytics.utils.checks import check_imshow, check_requirements |
||||
from ultralytics.solutions.object_counter import ObjectCounter # Import object counter class |
||||
from ultralytics.utils.plotting import Annotator |
||||
|
||||
check_requirements("shapely>=2.0.0") |
||||
|
||||
from shapely.geometry import LineString, Point, Polygon |
||||
|
||||
|
||||
class Heatmap: |
||||
class Heatmap(ObjectCounter): |
||||
"""A class to draw heatmaps in real-time video stream based on their tracks.""" |
||||
|
||||
def __init__( |
||||
self, |
||||
names, |
||||
colormap=cv2.COLORMAP_JET, |
||||
view_img=False, |
||||
view_in_counts=True, |
||||
view_out_counts=True, |
||||
count_reg_pts=None, |
||||
count_txt_color=(0, 0, 0), |
||||
count_bg_color=(255, 255, 255), |
||||
count_reg_color=(255, 0, 255), |
||||
region_thickness=5, |
||||
line_dist_thresh=15, |
||||
line_thickness=2, |
||||
shape="circle", |
||||
): |
||||
"""Initializes the heatmap class with default values for Visual, Image, track, count and heatmap parameters.""" |
||||
# Visual information |
||||
self.annotator = None |
||||
self.view_img = view_img |
||||
self.shape = shape |
||||
|
||||
self.initialized = False |
||||
self.names = names # Classes names |
||||
|
||||
# Image information |
||||
self.im0 = None |
||||
self.tf = line_thickness |
||||
self.view_in_counts = view_in_counts |
||||
self.view_out_counts = view_out_counts |
||||
|
||||
# Heatmap colormap and heatmap np array |
||||
self.colormap = colormap |
||||
self.heatmap = None |
||||
|
||||
# Predict/track information |
||||
self.boxes = [] |
||||
self.track_ids = [] |
||||
self.clss = [] |
||||
self.track_history = defaultdict(list) |
||||
|
||||
# Region & Line Information |
||||
self.counting_region = None |
||||
self.line_dist_thresh = line_dist_thresh |
||||
self.region_thickness = region_thickness |
||||
self.region_color = count_reg_color |
||||
|
||||
# Object Counting Information |
||||
self.in_counts = 0 |
||||
self.out_counts = 0 |
||||
self.count_ids = [] |
||||
self.class_wise_count = {} |
||||
self.count_txt_color = count_txt_color |
||||
self.count_bg_color = count_bg_color |
||||
self.cls_txtdisplay_gap = 50 |
||||
|
||||
# Check if environment supports imshow |
||||
self.env_check = check_imshow(warn=True) |
||||
|
||||
# Region and line selection |
||||
self.count_reg_pts = count_reg_pts |
||||
print(self.count_reg_pts) |
||||
if self.count_reg_pts is not None: |
||||
if len(self.count_reg_pts) == 2: |
||||
print("Line Counter Initiated.") |
||||
self.counting_region = LineString(self.count_reg_pts) |
||||
elif len(self.count_reg_pts) >= 3: |
||||
print("Polygon Counter Initiated.") |
||||
self.counting_region = Polygon(self.count_reg_pts) |
||||
else: |
||||
print("Invalid Region points provided, region_points must be 2 for lines or >= 3 for polygons.") |
||||
print("Using Line Counter Now") |
||||
self.counting_region = LineString(self.count_reg_pts) |
||||
|
||||
# Shape of heatmap, if not selected |
||||
if self.shape not in {"circle", "rect"}: |
||||
print("Unknown shape value provided, 'circle' & 'rect' supported") |
||||
print("Using Circular shape now") |
||||
self.shape = "circle" |
||||
|
||||
def extract_results(self, tracks): |
||||
""" |
||||
Extracts results from the provided data. |
||||
def __init__(self, **kwargs): |
||||
"""Initializes function for heatmap class with default values.""" |
||||
super().__init__(**kwargs) |
||||
|
||||
Args: |
||||
tracks (list): List of tracks obtained from the object tracking process. |
||||
""" |
||||
if tracks[0].boxes.id is not None: |
||||
self.boxes = tracks[0].boxes.xyxy.cpu() |
||||
self.clss = tracks[0].boxes.cls.tolist() |
||||
self.track_ids = tracks[0].boxes.id.int().tolist() |
||||
self.initialized = False # bool variable for heatmap initialization |
||||
if self.region is not None: # check if user provided the region coordinates |
||||
self.initialize_region() |
||||
|
||||
# store colormap |
||||
self.colormap = cv2.COLORMAP_PARULA if self.CFG["colormap"] is None else self.CFG["colormap"] |
||||
|
||||
def generate_heatmap(self, im0, tracks): |
||||
def heatmap_effect(self, box): |
||||
""" |
||||
Generate heatmap based on tracking data. |
||||
Efficient calculation of heatmap area and effect location for applying colormap. |
||||
|
||||
Args: |
||||
im0 (nd array): Image |
||||
tracks (list): List of tracks obtained from the object tracking process. |
||||
box (list): Bounding Box coordinates data [x0, y0, x1, y1] |
||||
""" |
||||
self.im0 = im0 |
||||
x0, y0, x1, y1 = map(int, box) |
||||
radius_squared = (min(x1 - x0, y1 - y0) // 2) ** 2 |
||||
|
||||
# Initialize heatmap only once |
||||
if not self.initialized: |
||||
self.heatmap = np.zeros((int(self.im0.shape[0]), int(self.im0.shape[1])), dtype=np.float32) |
||||
self.initialized = True |
||||
# Create a meshgrid with region of interest (ROI) for vectorized distance calculations |
||||
xv, yv = np.meshgrid(np.arange(x0, x1), np.arange(y0, y1)) |
||||
|
||||
self.heatmap *= 0.99 # decay factor |
||||
# Calculate squared distances from the center |
||||
dist_squared = (xv - ((x0 + x1) // 2)) ** 2 + (yv - ((y0 + y1) // 2)) ** 2 |
||||
|
||||
self.extract_results(tracks) |
||||
self.annotator = Annotator(self.im0, self.tf, None) |
||||
# Create a mask of points within the radius |
||||
within_radius = dist_squared <= radius_squared |
||||
|
||||
if self.track_ids: |
||||
# Draw counting region |
||||
if self.count_reg_pts is not None: |
||||
self.annotator.draw_region( |
||||
reg_pts=self.count_reg_pts, color=self.region_color, thickness=self.region_thickness |
||||
) |
||||
# Update only the values within the bounding box in a single vectorized operation |
||||
self.heatmap[y0:y1, x0:x1][within_radius] += 2 |
||||
|
||||
for box, cls, track_id in zip(self.boxes, self.clss, self.track_ids): |
||||
# Store class info |
||||
if self.names[cls] not in self.class_wise_count: |
||||
self.class_wise_count[self.names[cls]] = {"IN": 0, "OUT": 0} |
||||
|
||||
if self.shape == "circle": |
||||
center = (int((box[0] + box[2]) // 2), int((box[1] + box[3]) // 2)) |
||||
radius = min(int(box[2]) - int(box[0]), int(box[3]) - int(box[1])) // 2 |
||||
def generate_heatmap(self, im0): |
||||
""" |
||||
Generate heatmap for each frame using Ultralytics. |
||||
|
||||
y, x = np.ogrid[0 : self.heatmap.shape[0], 0 : self.heatmap.shape[1]] |
||||
mask = (x - center[0]) ** 2 + (y - center[1]) ** 2 <= radius**2 |
||||
Args: |
||||
im0 (ndarray): Input image array for processing |
||||
Returns: |
||||
im0 (ndarray): Processed image for further usage |
||||
""" |
||||
self.heatmap = np.zeros_like(im0, dtype=np.float32) * 0.99 if not self.initialized else self.heatmap |
||||
self.initialized = True # Initialize heatmap only once |
||||
|
||||
self.heatmap[int(box[1]) : int(box[3]), int(box[0]) : int(box[2])] += ( |
||||
2 * mask[int(box[1]) : int(box[3]), int(box[0]) : int(box[2])] |
||||
) |
||||
self.annotator = Annotator(im0, line_width=self.line_width) # Initialize annotator |
||||
self.extract_tracks(im0) # Extract tracks |
||||
|
||||
else: |
||||
self.heatmap[int(box[1]) : int(box[3]), int(box[0]) : int(box[2])] += 2 |
||||
# Iterate over bounding boxes, track ids and classes index |
||||
for box, track_id, cls in zip(self.boxes, self.track_ids, self.clss): |
||||
# Draw bounding box and counting region |
||||
self.heatmap_effect(box) |
||||
|
||||
# Store tracking hist |
||||
track_line = self.track_history[track_id] |
||||
track_line.append((float((box[0] + box[2]) / 2), float((box[1] + box[3]) / 2))) |
||||
if len(track_line) > 30: |
||||
track_line.pop(0) |
||||
if self.region is not None: |
||||
self.annotator.draw_region(reg_pts=self.region, color=(104, 0, 123), thickness=self.line_width * 2) |
||||
self.store_tracking_history(track_id, box) # Store track history |
||||
self.store_classwise_counts(cls) # store classwise counts in dict |
||||
|
||||
# Store tracking previous position and perform object counting |
||||
prev_position = self.track_history[track_id][-2] if len(self.track_history[track_id]) > 1 else None |
||||
self.count_objects(self.track_line, box, track_id, prev_position, cls) # Perform object counting |
||||
|
||||
if self.count_reg_pts is not None: |
||||
# Count objects in any polygon |
||||
if len(self.count_reg_pts) >= 3: |
||||
is_inside = self.counting_region.contains(Point(track_line[-1])) |
||||
|
||||
if prev_position is not None and is_inside and track_id not in self.count_ids: |
||||
self.count_ids.append(track_id) |
||||
|
||||
if (box[0] - prev_position[0]) * (self.counting_region.centroid.x - prev_position[0]) > 0: |
||||
self.in_counts += 1 |
||||
self.class_wise_count[self.names[cls]]["IN"] += 1 |
||||
else: |
||||
self.out_counts += 1 |
||||
self.class_wise_count[self.names[cls]]["OUT"] += 1 |
||||
|
||||
# Count objects using line |
||||
elif len(self.count_reg_pts) == 2: |
||||
if prev_position is not None and track_id not in self.count_ids: |
||||
distance = Point(track_line[-1]).distance(self.counting_region) |
||||
if distance < self.line_dist_thresh and track_id not in self.count_ids: |
||||
self.count_ids.append(track_id) |
||||
|
||||
if (box[0] - prev_position[0]) * ( |
||||
self.counting_region.centroid.x - prev_position[0] |
||||
) > 0: |
||||
self.in_counts += 1 |
||||
self.class_wise_count[self.names[cls]]["IN"] += 1 |
||||
else: |
||||
self.out_counts += 1 |
||||
self.class_wise_count[self.names[cls]]["OUT"] += 1 |
||||
|
||||
else: |
||||
for box, cls in zip(self.boxes, self.clss): |
||||
if self.shape == "circle": |
||||
center = (int((box[0] + box[2]) // 2), int((box[1] + box[3]) // 2)) |
||||
radius = min(int(box[2]) - int(box[0]), int(box[3]) - int(box[1])) // 2 |
||||
|
||||
y, x = np.ogrid[0 : self.heatmap.shape[0], 0 : self.heatmap.shape[1]] |
||||
mask = (x - center[0]) ** 2 + (y - center[1]) ** 2 <= radius**2 |
||||
|
||||
self.heatmap[int(box[1]) : int(box[3]), int(box[0]) : int(box[2])] += ( |
||||
2 * mask[int(box[1]) : int(box[3]), int(box[0]) : int(box[2])] |
||||
) |
||||
|
||||
else: |
||||
self.heatmap[int(box[1]) : int(box[3]), int(box[0]) : int(box[2])] += 2 |
||||
|
||||
if self.count_reg_pts is not None: |
||||
labels_dict = {} |
||||
|
||||
for key, value in self.class_wise_count.items(): |
||||
if value["IN"] != 0 or value["OUT"] != 0: |
||||
if not self.view_in_counts and not self.view_out_counts: |
||||
continue |
||||
elif not self.view_in_counts: |
||||
labels_dict[str.capitalize(key)] = f"OUT {value['OUT']}" |
||||
elif not self.view_out_counts: |
||||
labels_dict[str.capitalize(key)] = f"IN {value['IN']}" |
||||
else: |
||||
labels_dict[str.capitalize(key)] = f"IN {value['IN']} OUT {value['OUT']}" |
||||
|
||||
if labels_dict is not None: |
||||
self.annotator.display_analytics(self.im0, labels_dict, self.count_txt_color, self.count_bg_color, 10) |
||||
self.display_counts(im0) if self.region is not None else None # Display the counts on the frame |
||||
|
||||
# Normalize, apply colormap to heatmap and combine with original image |
||||
heatmap_normalized = cv2.normalize(self.heatmap, None, 0, 255, cv2.NORM_MINMAX) |
||||
heatmap_colored = cv2.applyColorMap(heatmap_normalized.astype(np.uint8), self.colormap) |
||||
self.im0 = cv2.addWeighted(self.im0, 0.5, heatmap_colored, 0.5, 0) |
||||
|
||||
if self.env_check and self.view_img: |
||||
self.display_frames() |
||||
|
||||
return self.im0 |
||||
|
||||
def display_frames(self): |
||||
"""Display frame.""" |
||||
cv2.imshow("Ultralytics Heatmap", self.im0) |
||||
|
||||
if cv2.waitKey(1) & 0xFF == ord("q"): |
||||
return |
||||
|
||||
|
||||
if __name__ == "__main__": |
||||
classes_names = {0: "person", 1: "car"} # example class names |
||||
heatmap = Heatmap(classes_names) |
||||
im0 = ( |
||||
im0 |
||||
if self.track_data.id is None |
||||
else cv2.addWeighted( |
||||
im0, |
||||
0.5, |
||||
cv2.applyColorMap( |
||||
cv2.normalize(self.heatmap, None, 0, 255, cv2.NORM_MINMAX).astype(np.uint8), self.colormap |
||||
), |
||||
0.5, |
||||
0, |
||||
) |
||||
) |
||||
|
||||
self.display_output(im0) # display output with base class function |
||||
return im0 # return output image for more usage |
||||
|
||||
Loading…
Reference in new issue