multimodalart's picture
Squashing commit
4450790 verified
import torch
from torchvision import transforms
import json
from PIL import Image, ImageDraw, ImageFont, ImageColor, ImageFilter, ImageChops
import numpy as np
from ..utility.utility import pil2tensor
import folder_paths
import io
import base64
from comfy.utils import common_upscale
def plot_coordinates_to_tensor(coordinates, height, width, bbox_height, bbox_width, size_multiplier, prompt):
import matplotlib
matplotlib.use('Agg')
from matplotlib.backends.backend_agg import FigureCanvasAgg as FigureCanvas
text_color = '#999999'
bg_color = '#353535'
matplotlib.pyplot.rcParams['text.color'] = text_color
fig, ax = matplotlib.pyplot.subplots(figsize=(width/100, height/100), dpi=100)
fig.patch.set_facecolor(bg_color)
ax.set_facecolor(bg_color)
ax.grid(color=text_color, linestyle='-', linewidth=0.5)
ax.set_xlabel('x', color=text_color)
ax.set_ylabel('y', color=text_color)
for text in ax.get_xticklabels() + ax.get_yticklabels():
text.set_color(text_color)
ax.set_title('position for: ' + prompt)
ax.set_xlabel('X Coordinate')
ax.set_ylabel('Y Coordinate')
#ax.legend().remove()
ax.set_xlim(0, width) # Set the x-axis to match the input latent width
ax.set_ylim(height, 0) # Set the y-axis to match the input latent height, with (0,0) at top-left
# Adjust the margins of the subplot
matplotlib.pyplot.subplots_adjust(left=0.08, right=0.95, bottom=0.05, top=0.95, wspace=0.2, hspace=0.2)
cmap = matplotlib.pyplot.get_cmap('rainbow')
image_batch = []
canvas = FigureCanvas(fig)
width, height = fig.get_size_inches() * fig.get_dpi()
# Draw a box at each coordinate
for i, ((x, y), size) in enumerate(zip(coordinates, size_multiplier)):
color_index = i / (len(coordinates) - 1)
color = cmap(color_index)
draw_height = bbox_height * size
draw_width = bbox_width * size
rect = matplotlib.patches.Rectangle((x - draw_width/2, y - draw_height/2), draw_width, draw_height,
linewidth=1, edgecolor=color, facecolor='none', alpha=0.5)
ax.add_patch(rect)
# Check if there is a next coordinate to draw an arrow to
if i < len(coordinates) - 1:
x1, y1 = coordinates[i]
x2, y2 = coordinates[i + 1]
ax.annotate("", xy=(x2, y2), xytext=(x1, y1),
arrowprops=dict(arrowstyle="->",
linestyle="-",
lw=1,
color=color,
mutation_scale=20))
canvas.draw()
image_np = np.frombuffer(canvas.tostring_rgb(), dtype='uint8').reshape(int(height), int(width), 3).copy()
image_tensor = torch.from_numpy(image_np).float() / 255.0
image_tensor = image_tensor.unsqueeze(0)
image_batch.append(image_tensor)
matplotlib.pyplot.close(fig)
image_batch_tensor = torch.cat(image_batch, dim=0)
return image_batch_tensor
class PlotCoordinates:
@classmethod
def INPUT_TYPES(s):
return {"required": {
"coordinates": ("STRING", {"forceInput": True}),
"text": ("STRING", {"default": 'title', "multiline": False}),
"width": ("INT", {"default": 512, "min": 8, "max": 4096, "step": 8}),
"height": ("INT", {"default": 512, "min": 8, "max": 4096, "step": 8}),
"bbox_width": ("INT", {"default": 128, "min": 8, "max": 4096, "step": 8}),
"bbox_height": ("INT", {"default": 128, "min": 8, "max": 4096, "step": 8}),
},
"optional": {"size_multiplier": ("FLOAT", {"default": [1.0], "forceInput": True})},
}
RETURN_TYPES = ("IMAGE", "INT", "INT", "INT", "INT",)
RETURN_NAMES = ("images", "width", "height", "bbox_width", "bbox_height",)
FUNCTION = "append"
CATEGORY = "KJNodes/experimental"
DESCRIPTION = """
Plots coordinates to sequence of images using Matplotlib.
"""
def append(self, coordinates, text, width, height, bbox_width, bbox_height, size_multiplier=[1.0]):
coordinates = json.loads(coordinates.replace("'", '"'))
coordinates = [(coord['x'], coord['y']) for coord in coordinates]
batch_size = len(coordinates)
if not size_multiplier or len(size_multiplier) != batch_size:
size_multiplier = [0] * batch_size
else:
size_multiplier = size_multiplier * (batch_size // len(size_multiplier)) + size_multiplier[:batch_size % len(size_multiplier)]
plot_image_tensor = plot_coordinates_to_tensor(coordinates, height, width, bbox_height, bbox_width, size_multiplier, text)
return (plot_image_tensor, width, height, bbox_width, bbox_height)
class SplineEditor:
@classmethod
def INPUT_TYPES(cls):
return {
"required": {
"points_store": ("STRING", {"multiline": False}),
"coordinates": ("STRING", {"multiline": False}),
"mask_width": ("INT", {"default": 512, "min": 8, "max": 4096, "step": 8}),
"mask_height": ("INT", {"default": 512, "min": 8, "max": 4096, "step": 8}),
"points_to_sample": ("INT", {"default": 16, "min": 2, "max": 1000, "step": 1}),
"sampling_method": (
[
'path',
'time',
'controlpoints'
],
{
"default": 'time'
}),
"interpolation": (
[
'cardinal',
'monotone',
'basis',
'linear',
'step-before',
'step-after',
'polar',
'polar-reverse',
],
{
"default": 'cardinal'
}),
"tension": ("FLOAT", {"default": 0.5, "min": 0.0, "max": 1.0, "step": 0.01}),
"repeat_output": ("INT", {"default": 1, "min": 1, "max": 4096, "step": 1}),
"float_output_type": (
[
'list',
'pandas series',
'tensor',
],
{
"default": 'list'
}),
},
"optional": {
"min_value": ("FLOAT", {"default": 0.0, "min": -10000.0, "max": 10000.0, "step": 0.01}),
"max_value": ("FLOAT", {"default": 1.0, "min": -10000.0, "max": 10000.0, "step": 0.01}),
"bg_image": ("IMAGE", ),
}
}
RETURN_TYPES = ("MASK", "STRING", "FLOAT", "INT", "STRING",)
RETURN_NAMES = ("mask", "coord_str", "float", "count", "normalized_str",)
FUNCTION = "splinedata"
CATEGORY = "KJNodes/weights"
DESCRIPTION = """
# WORK IN PROGRESS
Do not count on this as part of your workflow yet,
probably contains lots of bugs and stability is not
guaranteed!!
## Graphical editor to create values for various
## schedules and/or mask batches.
**Shift + click** to add control point at end.
**Ctrl + click** to add control point (subdivide) between two points.
**Right click on a point** to delete it.
Note that you can't delete from start/end.
Right click on canvas for context menu:
These are purely visual options, doesn't affect the output:
- Toggle handles visibility
- Display sample points: display the points to be returned.
**points_to_sample** value sets the number of samples
returned from the **drawn spline itself**, this is independent from the
actual control points, so the interpolation type matters.
sampling_method:
- time: samples along the time axis, used for schedules
- path: samples along the path itself, useful for coordinates
output types:
- mask batch
example compatible nodes: anything that takes masks
- list of floats
example compatible nodes: IPAdapter weights
- pandas series
example compatible nodes: anything that takes Fizz'
nodes Batch Value Schedule
- torch tensor
example compatible nodes: unknown
"""
def splinedata(self, mask_width, mask_height, coordinates, float_output_type, interpolation,
points_to_sample, sampling_method, points_store, tension, repeat_output,
min_value=0.0, max_value=1.0, bg_image=None):
coordinates = json.loads(coordinates)
normalized = []
normalized_y_values = []
for coord in coordinates:
coord['x'] = int(round(coord['x']))
coord['y'] = int(round(coord['y']))
norm_x = (1.0 - (coord['x'] / mask_height) - 0.0) * (max_value - min_value) + min_value
norm_y = (1.0 - (coord['y'] / mask_height) - 0.0) * (max_value - min_value) + min_value
normalized_y_values.append(norm_y)
normalized.append({'x':norm_x, 'y':norm_y})
if float_output_type == 'list':
out_floats = normalized_y_values * repeat_output
elif float_output_type == 'pandas series':
try:
import pandas as pd
except:
raise Exception("MaskOrImageToWeight: pandas is not installed. Please install pandas to use this output_type")
out_floats = pd.Series(normalized_y_values * repeat_output),
elif float_output_type == 'tensor':
out_floats = torch.tensor(normalized_y_values * repeat_output, dtype=torch.float32)
# Create a color map for grayscale intensities
color_map = lambda y: torch.full((mask_height, mask_width, 3), y, dtype=torch.float32)
# Create image tensors for each normalized y value
mask_tensors = [color_map(y) for y in normalized_y_values]
masks_out = torch.stack(mask_tensors)
masks_out = masks_out.repeat(repeat_output, 1, 1, 1)
masks_out = masks_out.mean(dim=-1)
if bg_image is None:
return (masks_out, json.dumps(coordinates), out_floats, len(out_floats) , json.dumps(normalized))
else:
transform = transforms.ToPILImage()
image = transform(bg_image[0].permute(2, 0, 1))
buffered = io.BytesIO()
image.save(buffered, format="JPEG", quality=75)
# Step 3: Encode the image bytes to a Base64 string
img_bytes = buffered.getvalue()
img_base64 = base64.b64encode(img_bytes).decode('utf-8')
return {
"ui": {"bg_image": [img_base64]},
"result":(masks_out, json.dumps(coordinates), out_floats, len(out_floats) , json.dumps(normalized))
}
class CreateShapeMaskOnPath:
RETURN_TYPES = ("MASK", "MASK",)
RETURN_NAMES = ("mask", "mask_inverted",)
FUNCTION = "createshapemask"
CATEGORY = "KJNodes/masking/generate"
DESCRIPTION = """
Creates a mask or batch of masks with the specified shape.
Locations are center locations.
"""
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"shape": (
[ 'circle',
'square',
'triangle',
],
{
"default": 'circle'
}),
"coordinates": ("STRING", {"forceInput": True}),
"frame_width": ("INT", {"default": 512,"min": 16, "max": 4096, "step": 1}),
"frame_height": ("INT", {"default": 512,"min": 16, "max": 4096, "step": 1}),
"shape_width": ("INT", {"default": 128,"min": 8, "max": 4096, "step": 1}),
"shape_height": ("INT", {"default": 128,"min": 8, "max": 4096, "step": 1}),
},
"optional": {
"size_multiplier": ("FLOAT", {"default": [1.0], "forceInput": True}),
}
}
def createshapemask(self, coordinates, frame_width, frame_height, shape_width, shape_height, shape, size_multiplier=[1.0]):
# Define the number of images in the batch
coordinates = coordinates.replace("'", '"')
coordinates = json.loads(coordinates)
batch_size = len(coordinates)
out = []
color = "white"
if not size_multiplier or len(size_multiplier) != batch_size:
size_multiplier = [0] * batch_size
else:
size_multiplier = size_multiplier * (batch_size // len(size_multiplier)) + size_multiplier[:batch_size % len(size_multiplier)]
for i, coord in enumerate(coordinates):
image = Image.new("RGB", (frame_width, frame_height), "black")
draw = ImageDraw.Draw(image)
# Calculate the size for this frame and ensure it's not less than 0
current_width = max(0, shape_width + i * size_multiplier[i])
current_height = max(0, shape_height + i * size_multiplier[i])
location_x = coord['x']
location_y = coord['y']
if shape == 'circle' or shape == 'square':
# Define the bounding box for the shape
left_up_point = (location_x - current_width // 2, location_y - current_height // 2)
right_down_point = (location_x + current_width // 2, location_y + current_height // 2)
two_points = [left_up_point, right_down_point]
if shape == 'circle':
draw.ellipse(two_points, fill=color)
elif shape == 'square':
draw.rectangle(two_points, fill=color)
elif shape == 'triangle':
# Define the points for the triangle
left_up_point = (location_x - current_width // 2, location_y + current_height // 2) # bottom left
right_down_point = (location_x + current_width // 2, location_y + current_height // 2) # bottom right
top_point = (location_x, location_y - current_height // 2) # top point
draw.polygon([top_point, left_up_point, right_down_point], fill=color)
image = pil2tensor(image)
mask = image[:, :, :, 0]
out.append(mask)
outstack = torch.cat(out, dim=0)
return (outstack, 1.0 - outstack,)
class CreateShapeImageOnPath:
RETURN_TYPES = ("IMAGE", "MASK",)
RETURN_NAMES = ("image","mask", )
FUNCTION = "createshapemask"
CATEGORY = "KJNodes/image"
DESCRIPTION = """
Creates an image or batch of images with the specified shape.
Locations are center locations.
"""
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"shape": (
[ 'circle',
'square',
'triangle',
],
{
"default": 'circle'
}),
"coordinates": ("STRING", {"forceInput": True}),
"frame_width": ("INT", {"default": 512,"min": 16, "max": 4096, "step": 1}),
"frame_height": ("INT", {"default": 512,"min": 16, "max": 4096, "step": 1}),
"shape_width": ("INT", {"default": 128,"min": 2, "max": 4096, "step": 1}),
"shape_height": ("INT", {"default": 128,"min": 2, "max": 4096, "step": 1}),
"shape_color": ("STRING", {"default": 'white'}),
"bg_color": ("STRING", {"default": 'black'}),
"blur_radius": ("FLOAT", {"default": 0.0, "min": 0.0, "max": 100, "step": 0.1}),
"intensity": ("FLOAT", {"default": 1.0, "min": 0.01, "max": 100.0, "step": 0.01}),
},
"optional": {
"size_multiplier": ("FLOAT", {"default": [1.0], "forceInput": True}),
"trailing": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 10.0, "step": 0.01}),
}
}
def createshapemask(self, coordinates, frame_width, frame_height, shape_width, shape_height, shape_color,
bg_color, blur_radius, shape, intensity, size_multiplier=[1.0], accumulate=False, trailing=1.0):
# Define the number of images in the batch
if len(coordinates) < 10:
coords_list = []
for coords in coordinates:
coords = json.loads(coords.replace("'", '"'))
coords_list.append(coords)
else:
coords = json.loads(coordinates.replace("'", '"'))
coords_list = [coords]
batch_size = len(coords_list[0])
images_list = []
masks_list = []
if not size_multiplier or len(size_multiplier) != batch_size:
size_multiplier = [0] * batch_size
else:
size_multiplier = size_multiplier * (batch_size // len(size_multiplier)) + size_multiplier[:batch_size % len(size_multiplier)]
previous_output = None
for i in range(batch_size):
image = Image.new("RGB", (frame_width, frame_height), bg_color)
draw = ImageDraw.Draw(image)
# Calculate the size for this frame and ensure it's not less than 0
current_width = max(0, shape_width + i * size_multiplier[i])
current_height = max(0, shape_height + i * size_multiplier[i])
for coords in coords_list:
location_x = coords[i]['x']
location_y = coords[i]['y']
if shape == 'circle' or shape == 'square':
# Define the bounding box for the shape
left_up_point = (location_x - current_width // 2, location_y - current_height // 2)
right_down_point = (location_x + current_width // 2, location_y + current_height // 2)
two_points = [left_up_point, right_down_point]
if shape == 'circle':
draw.ellipse(two_points, fill=shape_color)
elif shape == 'square':
draw.rectangle(two_points, fill=shape_color)
elif shape == 'triangle':
# Define the points for the triangle
left_up_point = (location_x - current_width // 2, location_y + current_height // 2) # bottom left
right_down_point = (location_x + current_width // 2, location_y + current_height // 2) # bottom right
top_point = (location_x, location_y - current_height // 2) # top point
draw.polygon([top_point, left_up_point, right_down_point], fill=shape_color)
if blur_radius != 0:
image = image.filter(ImageFilter.GaussianBlur(blur_radius))
# Blend the current image with the accumulated image
image = pil2tensor(image)
if trailing != 1.0 and previous_output is not None:
# Add the decayed previous output to the current frame
image += trailing * previous_output
image = image / image.max()
previous_output = image
image = image * intensity
mask = image[:, :, :, 0]
masks_list.append(mask)
images_list.append(image)
out_images = torch.cat(images_list, dim=0).cpu().float()
out_masks = torch.cat(masks_list, dim=0)
return (out_images, out_masks)
class CreateTextOnPath:
RETURN_TYPES = ("IMAGE", "MASK", "MASK",)
RETURN_NAMES = ("image", "mask", "mask_inverted",)
FUNCTION = "createtextmask"
CATEGORY = "KJNodes/masking/generate"
DESCRIPTION = """
Creates a mask or batch of masks with the specified text.
Locations are center locations.
"""
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"coordinates": ("STRING", {"forceInput": True}),
"text": ("STRING", {"default": 'text', "multiline": True}),
"frame_width": ("INT", {"default": 512,"min": 16, "max": 4096, "step": 1}),
"frame_height": ("INT", {"default": 512,"min": 16, "max": 4096, "step": 1}),
"font": (folder_paths.get_filename_list("kjnodes_fonts"), ),
"font_size": ("INT", {"default": 42}),
"alignment": (
[ 'left',
'center',
'right'
],
{"default": 'center'}
),
"text_color": ("STRING", {"default": 'white'}),
},
"optional": {
"size_multiplier": ("FLOAT", {"default": [1.0], "forceInput": True}),
}
}
def createtextmask(self, coordinates, frame_width, frame_height, font, font_size, text, text_color, alignment, size_multiplier=[1.0]):
coordinates = coordinates.replace("'", '"')
coordinates = json.loads(coordinates)
batch_size = len(coordinates)
mask_list = []
image_list = []
color = text_color
font_path = folder_paths.get_full_path("kjnodes_fonts", font)
if len(size_multiplier) != batch_size:
size_multiplier = size_multiplier * (batch_size // len(size_multiplier)) + size_multiplier[:batch_size % len(size_multiplier)]
for i, coord in enumerate(coordinates):
image = Image.new("RGB", (frame_width, frame_height), "black")
draw = ImageDraw.Draw(image)
lines = text.split('\n') # Split the text into lines
# Apply the size multiplier to the font size for this iteration
current_font_size = int(font_size * size_multiplier[i])
current_font = ImageFont.truetype(font_path, current_font_size)
line_heights = [current_font.getbbox(line)[3] for line in lines] # List of line heights
total_text_height = sum(line_heights) # Total height of text block
# Calculate the starting Y position to center the block of text
start_y = coord['y'] - total_text_height // 2
for j, line in enumerate(lines):
text_width, text_height = current_font.getbbox(line)[2], line_heights[j]
if alignment == 'left':
location_x = coord['x']
elif alignment == 'center':
location_x = int(coord['x'] - text_width // 2)
elif alignment == 'right':
location_x = int(coord['x'] - text_width)
location_y = int(start_y + sum(line_heights[:j]))
text_position = (location_x, location_y)
# Draw the text
try:
draw.text(text_position, line, fill=color, font=current_font, features=['-liga'])
except:
draw.text(text_position, line, fill=color, font=current_font)
image = pil2tensor(image)
non_black_pixels = (image > 0).any(dim=-1)
mask = non_black_pixels.to(image.dtype)
mask_list.append(mask)
image_list.append(image)
out_images = torch.cat(image_list, dim=0).cpu().float()
out_masks = torch.cat(mask_list, dim=0)
return (out_images, out_masks, 1.0 - out_masks,)
class CreateGradientFromCoords:
RETURN_TYPES = ("IMAGE", )
RETURN_NAMES = ("image", )
FUNCTION = "generate"
CATEGORY = "KJNodes/image"
DESCRIPTION = """
Creates a gradient image from coordinates.
"""
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"coordinates": ("STRING", {"forceInput": True}),
"frame_width": ("INT", {"default": 512,"min": 16, "max": 4096, "step": 1}),
"frame_height": ("INT", {"default": 512,"min": 16, "max": 4096, "step": 1}),
"start_color": ("STRING", {"default": 'white'}),
"end_color": ("STRING", {"default": 'black'}),
"multiplier": ("FLOAT", {"default": 1.0, "min": 0.01, "max": 100.0, "step": 0.01}),
},
}
def generate(self, coordinates, frame_width, frame_height, start_color, end_color, multiplier):
# Parse the coordinates
coordinates = json.loads(coordinates.replace("'", '"'))
# Create an image
image = Image.new("RGB", (frame_width, frame_height))
draw = ImageDraw.Draw(image)
# Extract start and end points for the gradient
start_coord = coordinates[0]
end_coord = coordinates[1]
start_color = ImageColor.getrgb(start_color)
end_color = ImageColor.getrgb(end_color)
# Calculate the gradient direction (vector)
gradient_direction = (end_coord['x'] - start_coord['x'], end_coord['y'] - start_coord['y'])
gradient_length = (gradient_direction[0] ** 2 + gradient_direction[1] ** 2) ** 0.5
# Iterate over each pixel in the image
for y in range(frame_height):
for x in range(frame_width):
# Calculate the projection of the point on the gradient line
point_vector = (x - start_coord['x'], y - start_coord['y'])
projection = (point_vector[0] * gradient_direction[0] + point_vector[1] * gradient_direction[1]) / gradient_length
projection = max(min(projection, gradient_length), 0) # Clamp the projection value
# Calculate the blend factor for the current pixel
blend = projection * multiplier / gradient_length
# Determine the color of the current pixel
color = (
int(start_color[0] + (end_color[0] - start_color[0]) * blend),
int(start_color[1] + (end_color[1] - start_color[1]) * blend),
int(start_color[2] + (end_color[2] - start_color[2]) * blend)
)
# Set the pixel color
draw.point((x, y), fill=color)
# Convert the PIL image to a tensor (assuming such a function exists in your context)
image_tensor = pil2tensor(image)
return (image_tensor,)
class GradientToFloat:
RETURN_TYPES = ("FLOAT", "FLOAT",)
RETURN_NAMES = ("float_x", "float_y", )
FUNCTION = "sample"
CATEGORY = "KJNodes/image"
DESCRIPTION = """
Calculates list of floats from image.
"""
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"image": ("IMAGE", ),
"steps": ("INT", {"default": 10, "min": 2, "max": 10000, "step": 1}),
},
}
def sample(self, image, steps):
# Assuming image is a tensor with shape [B, H, W, C]
B, H, W, C = image.shape
# Sample along the width axis (W)
w_intervals = torch.linspace(0, W - 1, steps=steps, dtype=torch.int64)
# Assuming we're sampling from the first batch and the first channel
w_sampled = image[0, :, w_intervals, 0]
# Sample along the height axis (H)
h_intervals = torch.linspace(0, H - 1, steps=steps, dtype=torch.int64)
# Assuming we're sampling from the first batch and the first channel
h_sampled = image[0, h_intervals, :, 0]
# Taking the mean across the height for width sampling, and across the width for height sampling
w_values = w_sampled.mean(dim=0).tolist()
h_values = h_sampled.mean(dim=1).tolist()
return (w_values, h_values)
class MaskOrImageToWeight:
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"output_type": (
[
'list',
'pandas series',
'tensor',
'string'
],
{
"default": 'list'
}),
},
"optional": {
"images": ("IMAGE",),
"masks": ("MASK",),
},
}
RETURN_TYPES = ("FLOAT", "STRING",)
FUNCTION = "execute"
CATEGORY = "KJNodes/weights"
DESCRIPTION = """
Gets the mean values from mask or image batch
and returns that as the selected output type.
"""
def execute(self, output_type, images=None, masks=None):
mean_values = []
if masks is not None and images is None:
for mask in masks:
mean_values.append(mask.mean().item())
elif masks is None and images is not None:
for image in images:
mean_values.append(image.mean().item())
elif masks is not None and images is not None:
raise Exception("MaskOrImageToWeight: Use either mask or image input only.")
# Convert mean_values to the specified output_type
if output_type == 'list':
out = mean_values
elif output_type == 'pandas series':
try:
import pandas as pd
except:
raise Exception("MaskOrImageToWeight: pandas is not installed. Please install pandas to use this output_type")
out = pd.Series(mean_values),
elif output_type == 'tensor':
out = torch.tensor(mean_values, dtype=torch.float32),
return (out, [str(value) for value in mean_values],)
class WeightScheduleConvert:
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"input_values": ("FLOAT", {"default": 0.0, "forceInput": True}),
"output_type": (
[
'match_input',
'list',
'pandas series',
'tensor',
],
{
"default": 'list'
}),
"invert": ("BOOLEAN", {"default": False}),
"repeat": ("INT", {"default": 1,"min": 1, "max": 255, "step": 1}),
},
"optional": {
"remap_to_frames": ("INT", {"default": 0}),
"interpolation_curve": ("FLOAT", {"forceInput": True}),
"remap_values": ("BOOLEAN", {"default": False}),
"remap_min": ("FLOAT", {"default": 0.0, "min": -100000, "max": 100000.0, "step": 0.01}),
"remap_max": ("FLOAT", {"default": 1.0, "min": -100000, "max": 100000.0, "step": 0.01}),
},
}
RETURN_TYPES = ("FLOAT", "STRING", "INT",)
FUNCTION = "execute"
CATEGORY = "KJNodes/weights"
DESCRIPTION = """
Converts different value lists/series to another type.
"""
def detect_input_type(self, input_values):
import pandas as pd
if isinstance(input_values, list):
return 'list'
elif isinstance(input_values, pd.Series):
return 'pandas series'
elif isinstance(input_values, torch.Tensor):
return 'tensor'
else:
raise ValueError("Unsupported input type")
def execute(self, input_values, output_type, invert, repeat, remap_to_frames=0, interpolation_curve=None, remap_min=0.0, remap_max=1.0, remap_values=False):
import pandas as pd
input_type = self.detect_input_type(input_values)
if input_type == 'pandas series':
float_values = input_values.tolist()
elif input_type == 'tensor':
float_values = input_values
else:
float_values = input_values
if invert:
float_values = [1 - value for value in float_values]
if interpolation_curve is not None:
interpolated_pattern = []
orig_float_values = float_values
for value in interpolation_curve:
min_val = min(orig_float_values)
max_val = max(orig_float_values)
# Normalize the values to [0, 1]
normalized_values = [(value - min_val) / (max_val - min_val) for value in orig_float_values]
# Interpolate the normalized values to the new frame count
remapped_float_values = np.interp(np.linspace(0, 1, int(remap_to_frames * value)), np.linspace(0, 1, len(normalized_values)), normalized_values).tolist()
interpolated_pattern.extend(remapped_float_values)
float_values = interpolated_pattern
else:
# Remap float_values to match target_frame_amount
if remap_to_frames > 0 and remap_to_frames != len(float_values):
min_val = min(float_values)
max_val = max(float_values)
# Normalize the values to [0, 1]
normalized_values = [(value - min_val) / (max_val - min_val) for value in float_values]
# Interpolate the normalized values to the new frame count
float_values = np.interp(np.linspace(0, 1, remap_to_frames), np.linspace(0, 1, len(normalized_values)), normalized_values).tolist()
float_values = float_values * repeat
if remap_values:
float_values = self.remap_values(float_values, remap_min, remap_max)
if output_type == 'list':
out = float_values,
elif output_type == 'pandas series':
out = pd.Series(float_values),
elif output_type == 'tensor':
if input_type == 'pandas series':
out = torch.tensor(float_values.values, dtype=torch.float32),
else:
out = torch.tensor(float_values, dtype=torch.float32),
elif output_type == 'match_input':
out = float_values,
return (out, [str(value) for value in float_values], [int(value) for value in float_values])
def remap_values(self, values, target_min, target_max):
# Determine the current range
current_min = min(values)
current_max = max(values)
current_range = current_max - current_min
# Determine the target range
target_range = target_max - target_min
# Perform the linear interpolation for each value
remapped_values = [(value - current_min) / current_range * target_range + target_min for value in values]
return remapped_values
class FloatToMask:
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"input_values": ("FLOAT", {"forceInput": True, "default": 0}),
"width": ("INT", {"default": 100, "min": 1}),
"height": ("INT", {"default": 100, "min": 1}),
},
}
RETURN_TYPES = ("MASK",)
FUNCTION = "execute"
CATEGORY = "KJNodes/masking/generate"
DESCRIPTION = """
Generates a batch of masks based on the input float values.
The batch size is determined by the length of the input float values.
Each mask is generated with the specified width and height.
"""
def execute(self, input_values, width, height):
import pandas as pd
# Ensure input_values is a list
if isinstance(input_values, (float, int)):
input_values = [input_values]
elif isinstance(input_values, pd.Series):
input_values = input_values.tolist()
elif isinstance(input_values, list) and all(isinstance(item, list) for item in input_values):
input_values = [item for sublist in input_values for item in sublist]
# Generate a batch of masks based on the input_values
masks = []
for value in input_values:
# Assuming value is a float between 0 and 1 representing the mask's intensity
mask = torch.ones((height, width), dtype=torch.float32) * value
masks.append(mask)
masks_out = torch.stack(masks, dim=0)
return(masks_out,)
class WeightScheduleExtend:
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"input_values_1": ("FLOAT", {"default": 0.0, "forceInput": True}),
"input_values_2": ("FLOAT", {"default": 0.0, "forceInput": True}),
"output_type": (
[
'match_input',
'list',
'pandas series',
'tensor',
],
{
"default": 'match_input'
}),
},
}
RETURN_TYPES = ("FLOAT",)
FUNCTION = "execute"
CATEGORY = "KJNodes/weights"
DESCRIPTION = """
Extends, and converts if needed, different value lists/series
"""
def detect_input_type(self, input_values):
import pandas as pd
if isinstance(input_values, list):
return 'list'
elif isinstance(input_values, pd.Series):
return 'pandas series'
elif isinstance(input_values, torch.Tensor):
return 'tensor'
else:
raise ValueError("Unsupported input type")
def execute(self, input_values_1, input_values_2, output_type):
import pandas as pd
input_type_1 = self.detect_input_type(input_values_1)
input_type_2 = self.detect_input_type(input_values_2)
# Convert input_values_2 to the same format as input_values_1 if they do not match
if not input_type_1 == input_type_2:
print("Converting input_values_2 to the same format as input_values_1")
if input_type_1 == 'pandas series':
# Convert input_values_2 to a pandas Series
float_values_2 = pd.Series(input_values_2)
elif input_type_1 == 'tensor':
# Convert input_values_2 to a tensor
float_values_2 = torch.tensor(input_values_2, dtype=torch.float32)
else:
print("Input types match, no conversion needed")
# If the types match, no conversion is needed
float_values_2 = input_values_2
float_values = input_values_1 + float_values_2
if output_type == 'list':
return float_values,
elif output_type == 'pandas series':
return pd.Series(float_values),
elif output_type == 'tensor':
if input_type_1 == 'pandas series':
return torch.tensor(float_values.values, dtype=torch.float32),
else:
return torch.tensor(float_values, dtype=torch.float32),
elif output_type == 'match_input':
return float_values,
else:
raise ValueError(f"Unsupported output_type: {output_type}")
class FloatToSigmas:
@classmethod
def INPUT_TYPES(s):
return {"required":
{
"float_list": ("FLOAT", {"default": 0.0, "forceInput": True}),
}
}
RETURN_TYPES = ("SIGMAS",)
RETURN_NAMES = ("SIGMAS",)
CATEGORY = "KJNodes/noise"
FUNCTION = "customsigmas"
DESCRIPTION = """
Creates a sigmas tensor from list of float values.
"""
def customsigmas(self, float_list):
return torch.tensor(float_list, dtype=torch.float32),
class SigmasToFloat:
@classmethod
def INPUT_TYPES(s):
return {"required":
{
"sigmas": ("SIGMAS",),
}
}
RETURN_TYPES = ("FLOAT",)
RETURN_NAMES = ("float",)
CATEGORY = "KJNodes/noise"
FUNCTION = "customsigmas"
DESCRIPTION = """
Creates a float list from sigmas tensors.
"""
def customsigmas(self, sigmas):
return sigmas.tolist(),
class GLIGENTextBoxApplyBatchCoords:
@classmethod
def INPUT_TYPES(s):
return {"required": {"conditioning_to": ("CONDITIONING", ),
"latents": ("LATENT", ),
"clip": ("CLIP", ),
"gligen_textbox_model": ("GLIGEN", ),
"coordinates": ("STRING", {"forceInput": True}),
"text": ("STRING", {"multiline": True}),
"width": ("INT", {"default": 128, "min": 8, "max": 4096, "step": 8}),
"height": ("INT", {"default": 128, "min": 8, "max": 4096, "step": 8}),
},
"optional": {"size_multiplier": ("FLOAT", {"default": [1.0], "forceInput": True})},
}
RETURN_TYPES = ("CONDITIONING", "IMAGE", )
RETURN_NAMES = ("conditioning", "coord_preview", )
FUNCTION = "append"
CATEGORY = "KJNodes/experimental"
DESCRIPTION = """
This node allows scheduling GLIGEN text box positions in a batch,
to be used with AnimateDiff-Evolved. Intended to pair with the
Spline Editor -node.
GLIGEN model can be downloaded through the Manage's "Install Models" menu.
Or directly from here:
https://huggingface.co/comfyanonymous/GLIGEN_pruned_safetensors/tree/main
Inputs:
- **latents** input is used to calculate batch size
- **clip** is your standard text encoder, use same as for the main prompt
- **gligen_textbox_model** connects to GLIGEN Loader
- **coordinates** takes a json string of points, directly compatible
with the spline editor node.
- **text** is the part of the prompt to set position for
- **width** and **height** are the size of the GLIGEN bounding box
Outputs:
- **conditioning** goes between to clip text encode and the sampler
- **coord_preview** is an optional preview of the coordinates and
bounding boxes.
"""
def append(self, latents, coordinates, conditioning_to, clip, gligen_textbox_model, text, width, height, size_multiplier=[1.0]):
coordinates = json.loads(coordinates.replace("'", '"'))
coordinates = [(coord['x'], coord['y']) for coord in coordinates]
batch_size = sum(tensor.size(0) for tensor in latents.values())
if len(coordinates) != batch_size:
print("GLIGENTextBoxApplyBatchCoords WARNING: The number of coordinates does not match the number of latents")
c = []
_, cond_pooled = clip.encode_from_tokens(clip.tokenize(text), return_pooled=True)
for t in conditioning_to:
n = [t[0], t[1].copy()]
position_params_batch = [[] for _ in range(batch_size)] # Initialize a list of empty lists for each batch item
if len(size_multiplier) != batch_size:
size_multiplier = size_multiplier * (batch_size // len(size_multiplier)) + size_multiplier[:batch_size % len(size_multiplier)]
for i in range(batch_size):
x_position, y_position = coordinates[i]
position_param = (cond_pooled, int((height // 8) * size_multiplier[i]), int((width // 8) * size_multiplier[i]), (y_position - height // 2) // 8, (x_position - width // 2) // 8)
position_params_batch[i].append(position_param) # Append position_param to the correct sublist
prev = []
if "gligen" in n[1]:
prev = n[1]['gligen'][2]
else:
prev = [[] for _ in range(batch_size)]
# Concatenate prev and position_params_batch, ensuring both are lists of lists
# and each sublist corresponds to a batch item
combined_position_params = [prev_item + batch_item for prev_item, batch_item in zip(prev, position_params_batch)]
n[1]['gligen'] = ("position_batched", gligen_textbox_model, combined_position_params)
c.append(n)
image_height = latents['samples'].shape[-2] * 8
image_width = latents['samples'].shape[-1] * 8
plot_image_tensor = plot_coordinates_to_tensor(coordinates, image_height, image_width, height, width, size_multiplier, text)
return (c, plot_image_tensor,)
class CreateInstanceDiffusionTracking:
RETURN_TYPES = ("TRACKING", "STRING", "INT", "INT", "INT", "INT",)
RETURN_NAMES = ("tracking", "prompt", "width", "height", "bbox_width", "bbox_height",)
FUNCTION = "tracking"
CATEGORY = "KJNodes/InstanceDiffusion"
DESCRIPTION = """
Creates tracking data to be used with InstanceDiffusion:
https://github.com/logtd/ComfyUI-InstanceDiffusion
InstanceDiffusion prompt format:
"class_id.class_name": "prompt",
for example:
"1.head": "((head))",
"""
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"coordinates": ("STRING", {"forceInput": True}),
"width": ("INT", {"default": 512,"min": 16, "max": 4096, "step": 1}),
"height": ("INT", {"default": 512,"min": 16, "max": 4096, "step": 1}),
"bbox_width": ("INT", {"default": 512,"min": 16, "max": 4096, "step": 1}),
"bbox_height": ("INT", {"default": 512,"min": 16, "max": 4096, "step": 1}),
"class_name": ("STRING", {"default": "class_name"}),
"class_id": ("INT", {"default": 0,"min": 0, "max": 255, "step": 1}),
"prompt": ("STRING", {"default": "prompt", "multiline": True}),
},
"optional": {
"size_multiplier": ("FLOAT", {"default": [1.0], "forceInput": True}),
"fit_in_frame": ("BOOLEAN", {"default": True}),
}
}
def tracking(self, coordinates, class_name, class_id, width, height, bbox_width, bbox_height, prompt, size_multiplier=[1.0], fit_in_frame=True):
# Define the number of images in the batch
coordinates = coordinates.replace("'", '"')
coordinates = json.loads(coordinates)
tracked = {}
tracked[class_name] = {}
batch_size = len(coordinates)
# Initialize a list to hold the coordinates for the current ID
id_coordinates = []
if not size_multiplier or len(size_multiplier) != batch_size:
size_multiplier = [0] * batch_size
else:
size_multiplier = size_multiplier * (batch_size // len(size_multiplier)) + size_multiplier[:batch_size % len(size_multiplier)]
for i, coord in enumerate(coordinates):
x = coord['x']
y = coord['y']
adjusted_bbox_width = bbox_width * size_multiplier[i]
adjusted_bbox_height = bbox_height * size_multiplier[i]
# Calculate the top left and bottom right coordinates
top_left_x = x - adjusted_bbox_width // 2
top_left_y = y - adjusted_bbox_height // 2
bottom_right_x = x + adjusted_bbox_width // 2
bottom_right_y = y + adjusted_bbox_height // 2
if fit_in_frame:
# Clip the coordinates to the frame boundaries
top_left_x = max(0, top_left_x)
top_left_y = max(0, top_left_y)
bottom_right_x = min(width, bottom_right_x)
bottom_right_y = min(height, bottom_right_y)
# Ensure width and height are positive
adjusted_bbox_width = max(1, bottom_right_x - top_left_x)
adjusted_bbox_height = max(1, bottom_right_y - top_left_y)
# Update the coordinates with the new width and height
bottom_right_x = top_left_x + adjusted_bbox_width
bottom_right_y = top_left_y + adjusted_bbox_height
# Append the top left and bottom right coordinates to the list for the current ID
id_coordinates.append([top_left_x, top_left_y, bottom_right_x, bottom_right_y, width, height])
class_id = int(class_id)
# Assign the list of coordinates to the specified ID within the class_id dictionary
tracked[class_name][class_id] = id_coordinates
prompt_string = ""
for class_name, class_data in tracked.items():
for class_id in class_data.keys():
class_id_str = str(class_id)
# Use the incoming prompt for each class name and ID
prompt_string += f'"{class_id_str}.{class_name}": "({prompt})",\n'
# Remove the last comma and newline
prompt_string = prompt_string.rstrip(",\n")
return (tracked, prompt_string, width, height, bbox_width, bbox_height)
class AppendInstanceDiffusionTracking:
RETURN_TYPES = ("TRACKING", "STRING",)
RETURN_NAMES = ("tracking", "prompt",)
FUNCTION = "append"
CATEGORY = "KJNodes/InstanceDiffusion"
DESCRIPTION = """
Appends tracking data to be used with InstanceDiffusion:
https://github.com/logtd/ComfyUI-InstanceDiffusion
"""
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"tracking_1": ("TRACKING", {"forceInput": True}),
"tracking_2": ("TRACKING", {"forceInput": True}),
},
"optional": {
"prompt_1": ("STRING", {"default": "", "forceInput": True}),
"prompt_2": ("STRING", {"default": "", "forceInput": True}),
}
}
def append(self, tracking_1, tracking_2, prompt_1="", prompt_2=""):
tracking_copy = tracking_1.copy()
# Check for existing class names and class IDs, and raise an error if they exist
for class_name, class_data in tracking_2.items():
if class_name not in tracking_copy:
tracking_copy[class_name] = class_data
else:
# If the class name exists, merge the class data from tracking_2 into tracking_copy
# This will add new class IDs under the same class name without raising an error
tracking_copy[class_name].update(class_data)
prompt_string = prompt_1 + "," + prompt_2
return (tracking_copy, prompt_string)
class InterpolateCoords:
RETURN_TYPES = ("STRING",)
RETURN_NAMES = ("coordinates",)
FUNCTION = "interpolate"
CATEGORY = "KJNodes/experimental"
DESCRIPTION = """
Interpolates coordinates based on a curve.
"""
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"coordinates": ("STRING", {"forceInput": True}),
"interpolation_curve": ("FLOAT", {"forceInput": True}),
},
}
def interpolate(self, coordinates, interpolation_curve):
# Parse the JSON string to get the list of coordinates
coordinates = json.loads(coordinates.replace("'", '"'))
# Convert the list of dictionaries to a list of (x, y) tuples for easier processing
coordinates = [(coord['x'], coord['y']) for coord in coordinates]
# Calculate the total length of the original path
path_length = sum(np.linalg.norm(np.array(coordinates[i]) - np.array(coordinates[i-1]))
for i in range(1, len(coordinates)))
# Initialize variables for interpolation
interpolated_coords = []
current_length = 0
current_index = 0
# Iterate over the normalized curve
for normalized_length in interpolation_curve:
target_length = normalized_length * path_length # Convert to the original scale
while current_index < len(coordinates) - 1:
segment_start, segment_end = np.array(coordinates[current_index]), np.array(coordinates[current_index + 1])
segment_length = np.linalg.norm(segment_end - segment_start)
if current_length + segment_length >= target_length:
break
current_length += segment_length
current_index += 1
# Interpolate between the last two points
if current_index < len(coordinates) - 1:
p1, p2 = np.array(coordinates[current_index]), np.array(coordinates[current_index + 1])
segment_length = np.linalg.norm(p2 - p1)
if segment_length > 0:
t = (target_length - current_length) / segment_length
interpolated_point = p1 + t * (p2 - p1)
interpolated_coords.append(interpolated_point.tolist())
else:
interpolated_coords.append(p1.tolist())
else:
# If the target_length is at or beyond the end of the path, add the last coordinate
interpolated_coords.append(coordinates[-1])
# Convert back to string format if necessary
interpolated_coords_str = "[" + ", ".join([f"{{'x': {round(coord[0])}, 'y': {round(coord[1])}}}" for coord in interpolated_coords]) + "]"
print(interpolated_coords_str)
return (interpolated_coords_str,)
class DrawInstanceDiffusionTracking:
RETURN_TYPES = ("IMAGE",)
RETURN_NAMES = ("image", )
FUNCTION = "draw"
CATEGORY = "KJNodes/InstanceDiffusion"
DESCRIPTION = """
Draws the tracking data from
CreateInstanceDiffusionTracking -node.
"""
@classmethod
def INPUT_TYPES(s):
return {
"required": {
"image": ("IMAGE", ),
"tracking": ("TRACKING", {"forceInput": True}),
"box_line_width": ("INT", {"default": 2, "min": 1, "max": 10, "step": 1}),
"draw_text": ("BOOLEAN", {"default": True}),
"font": (folder_paths.get_filename_list("kjnodes_fonts"), ),
"font_size": ("INT", {"default": 20}),
},
}
def draw(self, image, tracking, box_line_width, draw_text, font, font_size):
import matplotlib.cm as cm
modified_images = []
colormap = cm.get_cmap('rainbow', len(tracking))
if draw_text:
font_path = folder_paths.get_full_path("kjnodes_fonts", font)
font = ImageFont.truetype(font_path, font_size)
# Iterate over each image in the batch
for i in range(image.shape[0]):
# Extract the current image and convert it to a PIL image
current_image = image[i, :, :, :].permute(2, 0, 1)
pil_image = transforms.ToPILImage()(current_image)
draw = ImageDraw.Draw(pil_image)
# Iterate over the bounding boxes for the current image
for j, (class_name, class_data) in enumerate(tracking.items()):
for class_id, bbox_list in class_data.items():
# Check if the current index is within the bounds of the bbox_list
if i < len(bbox_list):
bbox = bbox_list[i]
# Ensure bbox is a list or tuple before unpacking
if isinstance(bbox, (list, tuple)):
x1, y1, x2, y2, _, _ = bbox
# Convert coordinates to integers
x1, y1, x2, y2 = int(x1), int(y1), int(x2), int(y2)
# Generate a color from the rainbow colormap
color = tuple(int(255 * x) for x in colormap(j / len(tracking)))[:3]
# Draw the bounding box on the image with the generated color
draw.rectangle([x1, y1, x2, y2], outline=color, width=box_line_width)
if draw_text:
# Draw the class name and ID as text above the box with the generated color
text = f"{class_id}.{class_name}"
# Calculate the width and height of the text
_, _, text_width, text_height = draw.textbbox((0, 0), text=text, font=font)
# Position the text above the top-left corner of the box
text_position = (x1, y1 - text_height)
draw.text(text_position, text, fill=color, font=font)
else:
print(f"Unexpected data type for bbox: {type(bbox)}")
# Convert the drawn image back to a torch tensor and adjust back to (H, W, C)
modified_image_tensor = transforms.ToTensor()(pil_image).permute(1, 2, 0)
modified_images.append(modified_image_tensor)
# Stack the modified images back into a batch
image_tensor_batch = torch.stack(modified_images).cpu().float()
return image_tensor_batch,
class PointsEditor:
@classmethod
def INPUT_TYPES(cls):
return {
"required": {
"points_store": ("STRING", {"multiline": False}),
"coordinates": ("STRING", {"multiline": False}),
"neg_coordinates": ("STRING", {"multiline": False}),
"bbox_store": ("STRING", {"multiline": False}),
"bboxes": ("STRING", {"multiline": False}),
"bbox_format": (
[
'xyxy',
'xywh',
],
),
"width": ("INT", {"default": 512, "min": 8, "max": 4096, "step": 8}),
"height": ("INT", {"default": 512, "min": 8, "max": 4096, "step": 8}),
"normalize": ("BOOLEAN", {"default": False}),
},
"optional": {
"bg_image": ("IMAGE", ),
},
}
RETURN_TYPES = ("STRING", "STRING", "BBOX", "MASK", "IMAGE")
RETURN_NAMES = ("positive_coords", "negative_coords", "bbox", "bbox_mask", "cropped_image")
FUNCTION = "pointdata"
CATEGORY = "KJNodes/experimental"
DESCRIPTION = """
# WORK IN PROGRESS
Do not count on this as part of your workflow yet,
probably contains lots of bugs and stability is not
guaranteed!!
## Graphical editor to create coordinates
**Shift + click** to add a positive (green) point.
**Shift + right click** to add a negative (red) point.
**Ctrl + click** to draw a box.
**Right click on a point** to delete it.
Note that you can't delete from start/end of the points array.
To add an image select the node and copy/paste or drag in the image.
Or from the bg_image input on queue (first frame of the batch).
**THE IMAGE IS SAVED TO THE NODE AND WORKFLOW METADATA**
you can clear the image from the context menu by right clicking on the canvas
"""
def pointdata(self, points_store, bbox_store, width, height, coordinates, neg_coordinates, normalize, bboxes, bbox_format="xyxy", bg_image=None):
coordinates = json.loads(coordinates)
pos_coordinates = []
for coord in coordinates:
coord['x'] = int(round(coord['x']))
coord['y'] = int(round(coord['y']))
if normalize:
norm_x = coord['x'] / width
norm_y = coord['y'] / height
pos_coordinates.append({'x': norm_x, 'y': norm_y})
else:
pos_coordinates.append({'x': coord['x'], 'y': coord['y']})
if neg_coordinates:
coordinates = json.loads(neg_coordinates)
neg_coordinates = []
for coord in coordinates:
coord['x'] = int(round(coord['x']))
coord['y'] = int(round(coord['y']))
if normalize:
norm_x = coord['x'] / width
norm_y = coord['y'] / height
neg_coordinates.append({'x': norm_x, 'y': norm_y})
else:
neg_coordinates.append({'x': coord['x'], 'y': coord['y']})
# Create a blank mask
mask = np.zeros((height, width), dtype=np.uint8)
bboxes = json.loads(bboxes)
print(bboxes)
valid_bboxes = []
for bbox in bboxes:
if (bbox.get("startX") is None or
bbox.get("startY") is None or
bbox.get("endX") is None or
bbox.get("endY") is None):
continue # Skip this bounding box if any value is None
else:
# Ensure that endX and endY are greater than startX and startY
x_min = min(int(bbox["startX"]), int(bbox["endX"]))
y_min = min(int(bbox["startY"]), int(bbox["endY"]))
x_max = max(int(bbox["startX"]), int(bbox["endX"]))
y_max = max(int(bbox["startY"]), int(bbox["endY"]))
valid_bboxes.append((x_min, y_min, x_max, y_max))
bboxes_xyxy = []
for bbox in valid_bboxes:
x_min, y_min, x_max, y_max = bbox
bboxes_xyxy.append((x_min, y_min, x_max, y_max))
mask[y_min:y_max, x_min:x_max] = 1 # Fill the bounding box area with 1s
if bbox_format == "xywh":
bboxes_xywh = []
for bbox in valid_bboxes:
x_min, y_min, x_max, y_max = bbox
width = x_max - x_min
height = y_max - y_min
bboxes_xywh.append((x_min, y_min, width, height))
bboxes = bboxes_xywh
else:
bboxes = bboxes_xyxy
mask_tensor = torch.from_numpy(mask)
mask_tensor = mask_tensor.unsqueeze(0).float().cpu()
if bg_image is not None and len(valid_bboxes) > 0:
x_min, y_min, x_max, y_max = bboxes[0]
cropped_image = bg_image[:, y_min:y_max, x_min:x_max, :]
elif bg_image is not None:
cropped_image = bg_image
if bg_image is None:
return (json.dumps(pos_coordinates), json.dumps(neg_coordinates), bboxes, mask_tensor)
else:
transform = transforms.ToPILImage()
image = transform(bg_image[0].permute(2, 0, 1))
buffered = io.BytesIO()
image.save(buffered, format="JPEG", quality=75)
# Step 3: Encode the image bytes to a Base64 string
img_bytes = buffered.getvalue()
img_base64 = base64.b64encode(img_bytes).decode('utf-8')
return {
"ui": {"bg_image": [img_base64]},
"result": (json.dumps(pos_coordinates), json.dumps(neg_coordinates), bboxes, mask_tensor, cropped_image)
}