From a2a126e7551792f2543b220920ba9ea331ef3a0a Mon Sep 17 00:00:00 2001
From: mhby1g21 <mhby1g21@soton.ac.uk>
Date: Wed, 16 Oct 2024 01:14:07 +0100
Subject: [PATCH] dynamic dir changes and no need to use path.txt no more,
added new folder to gitignore
---
.gitignore | 2 +-
combine_img.py | 86 +++++++++++++++++++-------------------------------
split_img.py | 62 +++++++++++++++---------------------
ui.py | 12 +++++--
4 files changed, 68 insertions(+), 94 deletions(-)
diff --git a/.gitignore b/.gitignore
index 36b3b53..98ff992 100644
--- a/.gitignore
+++ b/.gitignore
@@ -6,7 +6,7 @@ data/matbase/masks/*
# intermediate files
output/
-split_output/
+cubemap_faces/
# Byte-compiled / optimized / DLL files
__pycache__/
diff --git a/combine_img.py b/combine_img.py
index 4e02ebd..691fdeb 100644
--- a/combine_img.py
+++ b/combine_img.py
@@ -1,6 +1,7 @@
import numpy as np
import imageio
import os
+import sys
from ui import get_res
#makes sure value of x stays within the range of min and max value to prevent out of bound values accses for the image
@@ -9,29 +10,11 @@ def clamp(x, min_val, max_val):
#clamps the floating point coordinates to nearest integer value, copies pixel value from the image to the nearest int coord
def nearest_neighbour_interpolation(img, x, y):
- h, w, _ = img.shape
+ # Modified to handle both 2D and 3D images
+ h, w = img.shape[:2]
x, y = clamp(int(x), 0, w-1), clamp(int(y), 0, h-1)
return img[y, x]
-
-# def orientation_to_face(x, y, z):
-# abs_x, abs_y, abs_z = abs(x), abs(y), abs(z)
-# if abs_x >= abs_y and abs_x >= abs_z:
-# if x > 0:
-# return 'front', -y / abs_x, -z / abs_x
-# else:
-# return 'back', y / abs_x, -z / abs_x
-# elif abs_y >= abs_x and abs_y >= abs_z:
-# if y > 0:
-# return 'right', -x / abs_y, -z / abs_y
-# else:
-# return 'left', x / abs_y, -z / abs_y
-# else:
-# if z > 0:
-# return 'top', x / abs_z, y / abs_z
-# else:
-# return 'bottom', -x / abs_z, y / abs_z
-
#maps the 3d coords for cube faces to 2d coords on that cube face
#finds out which cube face corresponds to the current pixel for which the 3d coords are calculated and calcs the normalised 2d coords on that face
def orientation_to_face(x, y, z):
@@ -54,7 +37,14 @@ def orientation_to_face(x, y, z):
#converts cube maps into omnidirectional image
def cubemap_to_omnidirectional(cube_faces, out_width, out_height):
- omnidirectional = np.zeros((out_height, out_width, 3), dtype=np.uint8)
+ # Determine if the input is grayscale or RGB
+ is_grayscale = len(cube_faces[list(cube_faces.keys())[0]].shape) == 2
+
+ # Create the appropriate output array based on input type
+ if is_grayscale:
+ omnidirectional = np.zeros((out_height, out_width), dtype=np.uint8)
+ else:
+ omnidirectional = np.zeros((out_height, out_width, 3), dtype=np.uint8)
#iterates through the pixels in o/p image, for each pixel calulates spherical coord used to map 2d pixel loc to 3d points on a sphere
#which are then converted to 3d cartesian coord to find which face of cube map current pixel corresponds to
@@ -79,23 +69,27 @@ def cubemap_to_omnidirectional(cube_faces, out_width, out_height):
return omnidirectional
-# if __name__ == "__main__":
-# # Load the cubemap images
-# cube_faces_dir = input("Enter the directory containing the cubemap images: ").strip()
-# faces = ["right", "left", "top", "bottom", "front", "back"]
-# cube_faces = {}
-
-# for face in faces:
-# cube_faces[face] = imageio.imread(os.path.join(cube_faces_dir, f"{face}.jpg"))
+if __name__ == "__main__":
+ # Get the directory of the current script
+ script_dir = os.path.dirname(os.path.abspath(__file__))
+
+ # Default paths relative to the script directory
+ default_cube_faces_dir = os.path.join(script_dir, "output", "cubemap_faces")
+ default_output_dir = os.path.join(script_dir, "..", "edgenet-360", "Data", "Input")
+ rgb_png_path = os.path.join(script_dir, "..", "scripts", "360monodepthexecution", "rgb.jpg")
+ # Use command-line arguments for custom paths if provided
+ if len(sys.argv) > 2:
+ cube_faces_dir = sys.argv[1]
+ output_dir = sys.argv[2]
+ else:
+ cube_faces_dir = default_cube_faces_dir
+ output_dir = default_output_dir
+ # Ensure output directory exists
+ os.makedirs(output_dir, exist_ok=True)
-if __name__ == "__main__":
# Load the cubemap images
- #cube_faces_dir = input("Enter the directory containing the cubemap images: ").strip()
- cube_faces_dir = "C:\Project\AVVR-Pipeline-Internship\material_recognition\Dynamic-Backward-Attention-Transformer\output\split_output"
-
- #faces = ["right", "left", "top", "bottom", "front", "back"]
faces = ["rightrgb", "leftrgb", "toprgb", "bottomrgb", "frontrgb", "backrgb"]
cube_faces = {}
@@ -103,39 +97,23 @@ if __name__ == "__main__":
image_path = os.path.join(cube_faces_dir, f"{face}.png")
image_data = imageio.imread(image_path)
- #rotate top and bottom face by 90 deg
- # if face in ["top", "bottom"]:
- # image_data = np.rot90(image_data, 1)
-
- # #flip the top, bottom, front and back faces in horizontal direction
- # if face not in ["left", "right"]:
- # image_data = image_data[:, ::-1]
-
if face in ["toprgb", "bottomrgb"]:
image_data = np.rot90(image_data, 1)
if face not in ["leftrgb", "rightrgb"]:
image_data = image_data[:, ::-1]
-
cube_faces[face] = image_data
-
- # output_width = int(input("Enter output omnidirectional width: "))
- # output_height = int(input("Enter output omnidirectional height: "))
- with open('path.txt', 'r') as file:
- input_path = file.readline()
- print(f'path = {input_path}')
- os.remove('path.txt')
- height, width = get_res(input_path)
- print(height, width)
+ # Use the rgb.png file from edgenet-360 folder for dimensions
+ height, width = get_res(rgb_png_path)
+ print(f"Using dimensions from {rgb_png_path}: {width}x{height}")
output_width = width
output_height = height
- #print(f"height: {height}, width: {width}")
omnidirectional_img = cubemap_to_omnidirectional(cube_faces, output_width, output_height)
- output_path = "C:\Project\AVVR-Pipeline-Internship\edgenet360\Data\Input\material.png"
+ output_path = os.path.join(output_dir, "material.png")
imageio.v2.imsave(output_path, omnidirectional_img)
print(f"Omnidirectional image saved to {output_path}")
diff --git a/split_img.py b/split_img.py
index 83459dd..969a526 100644
--- a/split_img.py
+++ b/split_img.py
@@ -1,6 +1,5 @@
###### This code has been Referenced from: https://github.com/jaxry/panorama-to-cubemap/blob/gh-pages/convert.js
-
import numpy as np
import imageio
import os
@@ -16,7 +15,7 @@ def mod(x, n):
return ((x % n) + n) % n
#clamps the floating point coordinates to nearest integer value, copies pixel value from the image to the nearest int coord
-def nearest_neigbour_interpolation(img, x, y):
+def nearest_neighbour_interpolation(img, x, y):
h, w, _ = img.shape
x, y = clamp(int(x), 0, w-1), clamp(int(y), 0, h-1)
return img[y, x]
@@ -49,56 +48,45 @@ def face_rendering(img, face, face_size):
longitude = mod(np.arctan2(out[1], out[0]), 2 * np.pi)
latitude = np.arccos(out[2] / r)
s_x, s_y = img.shape[1] * longitude / (2 * np.pi) - 0.5, img.shape[0] * latitude / np.pi - 0.5
- out_face[y, x] = nearest_neigbour_interpolation(img, s_x, s_y)
+ out_face[y, x] = nearest_neighbour_interpolation(img, s_x, s_y)
return out_face
#generates 6 cube faces
-def generate_cube_faces(input_path, output_path="cube_faces_output"):
-
+def generate_cube_faces(input_path, output_path):
img = imageio.imread(input_path)
face_size = 512 #each face o/p image will be 512x512
faces = ["right", "left", "top", "bottom", "front", "back"]
- results = {}
for face in faces:
- results[face] = face_rendering(img, face, face_size)
+ face_img = face_rendering(img, face, face_size)
face_output_path = os.path.join(output_path, f"{face}.png")
- imageio.imsave(face_output_path, results[face])
+ imageio.imsave(face_output_path, face_img)
print(f"Saved {face} face to {face_output_path}")
-
-
-
-
-# if __name__ == "__main__":
-# input_path = select_image()
-# height, width = get_res(input_path)
-# print("width: , Height:", width, height)
-# #output_path = input("Enter output directory: ").strip()
-# #if not output_path:
-# output_path = "C:\Project\AVVR-Pipeline-Internship\material_recognition\Dynamic-Backward-Attention-Transformer\split_output"
-# if not os.path.exists(output_path):
-# os.makedirs(output_path)
-# generate_cube_faces(input_path, output_path)
-
-
-#input_path = select_image()
-input_path = sys.argv[1]
-with open('path.txt', 'w') as file:
- file.write(input_path)
- file.close()
-
-#output_path = input("Enter output directory: ").strip()
-#if not output_path:
-output_path = "C:\Project\AVVR-Pipeline-Internship\material_recognition\Dynamic-Backward-Attention-Transformer\split_output"
-if not os.path.exists(output_path):
- os.makedirs(output_path)
-generate_cube_faces(input_path, output_path)
-
+if __name__ == "__main__":
+ # Get the directory of the current script
+ script_dir = os.path.dirname(os.path.abspath(__file__))
+ # Default output path relative to the script directory
+ default_output_path = os.path.join(script_dir, "cubemap_faces")
+ # Use command-line arguments if provided
+ if len(sys.argv) > 1:
+ input_path = sys.argv[1]
+ if len(sys.argv) > 2:
+ output_path = sys.argv[2]
+ else:
+ output_path = default_output_path
+ else:
+ input_path = select_image()
+ output_path = default_output_path
+ # Ensure output directory exists
+ os.makedirs(output_path, exist_ok=True)
+ # Generate cube faces
+ generate_cube_faces(input_path, output_path)
+ print(f"Cubemap faces saved to: {output_path}")
\ No newline at end of file
diff --git a/ui.py b/ui.py
index 6cd25e6..e40408f 100644
--- a/ui.py
+++ b/ui.py
@@ -1,9 +1,9 @@
import tkinter as tk
from tkinter import filedialog
import cv2
+import os
def select_image():
-
root = tk.Tk()
root.withdraw()
@@ -16,7 +16,15 @@ def select_image():
def get_res(path):
img = cv2.imread(path)
height, width, _ = img.shape
- cv2.imwrite('C:\Project\AVVR-Pipeline-Internship\edgenet360\Data\Input\\rgb.png', img)
+
+ # Construct the output path dynamically
+ script_dir = os.path.dirname(os.path.abspath(__file__))
+ output_dir = os.path.join(script_dir, "..", "edgenet-360", "Data", "Input")
+ os.makedirs(output_dir, exist_ok=True)
+ output_path = os.path.join(output_dir, "rgb.png")
+
+ cv2.imwrite(output_path, img)
+ print(f"RGB image saved to: {output_path}")
return height, width
--
GitLab