Solve small issues with preprocessing + visualization

This commit is contained in:
MrKarimiD committed 2025-06-05 13:51:35 -04:00
1 parent e3db99e017
commit ccad510a60
7 files changed
+251 -7

No files matched your search

+21
View File
@@ -2,6 +2,8 @@ import subprocess
import argparse
import os
import shutil
import sys
if __name__ == '__main__':
argparser = argparse.ArgumentParser()
@@ -20,16 +22,35 @@ if __name__ == '__main__':
metrics_dir = os.path.join(args.input_dir, "metrics")
renders_dir = os.path.join(metrics_dir, "images")
renders_well = os.path.join(renders_dir, args.metric_name)
if os.path.exists(renders_well):
shutil.rmtree(renders_well)
renders_fast = os.path.join(renders_dir, args.metric_name + "_fast")
if os.path.exists(renders_fast):
shutil.rmtree(renders_fast)
renders_lin = os.path.join(renders_dir, args.metric_name + "_lin")
if os.path.exists(renders_lin):
shutil.rmtree(renders_lin)
renders_lin_HDR = os.path.join(renders_dir, args.metric_name + "_lin_HDR")
if os.path.exists(renders_lin_HDR):
shutil.rmtree(renders_lin_HDR)
lin_renders = os.path.join(renders_dir, args.metric_name + "_lin_renders")
if os.path.exists(lin_renders):
shutil.rmtree(lin_renders)
gt_dir = os.path.join(args.input_dir, "GT")
gt_hdr = os.path.join(gt_dir, "GT_hdr")
if os.path.exists(gt_hdr):
shutil.rmtree(gt_hdr)
gt_exr = os.path.join(gt_dir, "GT_exr")
gt_exr_renders = os.path.join(gt_dir, "GT_exr_renders")
if os.path.exists(gt_exr_renders):
shutil.rmtree(gt_exr_renders)
# 1. Get ground truth transformations:
print("Running ns-process-data to get GT transformations...")
@@ -2,6 +2,8 @@ import subprocess
import argparse
import os
import shutil
import sys
if __name__ == '__main__':
argparser = argparse.ArgumentParser()
@@ -20,16 +22,35 @@ if __name__ == '__main__':
metrics_dir = os.path.join(args.input_dir, "metrics")
renders_dir = os.path.join(metrics_dir, "images")
renders_well = os.path.join(renders_dir, args.metric_name)
if os.path.exists(renders_well):
shutil.rmtree(renders_well)
renders_fast = os.path.join(renders_dir, args.metric_name + "_fast")
if os.path.exists(renders_fast):
shutil.rmtree(renders_fast)
renders_lin = os.path.join(renders_dir, args.metric_name + "_lin")
if os.path.exists(renders_lin):
shutil.rmtree(renders_lin)
renders_lin_HDR = os.path.join(renders_dir, args.metric_name + "_lin_HDR")
if os.path.exists(renders_lin_HDR):
shutil.rmtree(renders_lin_HDR)
lin_renders = os.path.join(renders_dir, args.metric_name + "_lin_renders")
if os.path.exists(lin_renders):
shutil.rmtree(lin_renders)
gt_dir = os.path.join(args.input_dir, "GT")
gt_hdr = os.path.join(gt_dir, "GT_hdr")
if os.path.exists(gt_hdr):
shutil.rmtree(gt_hdr)
gt_exr = os.path.join(gt_dir, "GT_exr")
gt_exr_renders = os.path.join(gt_dir, "GT_exr_renders")
if os.path.exists(gt_exr_renders):
shutil.rmtree(gt_exr_renders)
# 1. Get ground truth transformations:
print("Running ns-process-data to get GT transformations...")
@@ -156,3 +177,38 @@ if __name__ == '__main__':
"--hdr_dir", gt_exr,
"--output_dir", gt_hdr
], check=True)
# 9. Convert EXR panos to LDR format:
print("Converting EXR panos to LDR visualization...")
subprocess.run([
"python", "lantern_scripts/tonemap.py",
"--data_dir", renders_lin,
"--out_dir", renders_lin
], check=True)
subprocess.run([
"python", "lantern_scripts/tonemap.py",
"--data_dir", gt_exr,
"--out_dir", gt_exr
], check=True)
subprocess.run([
"python", "lantern_scripts/tonemap.py",
"--data_dir", lin_renders,
"--out_dir", lin_renders
], check=True)
subprocess.run([
"python", "lantern_scripts/tonemap.py",
"--data_dir", gt_exr_renders,
"--out_dir", gt_exr_renders
], check=True)
# 9. Convert EXR panos to LDR format:
print("Preparing the HTML visualization...")
subprocess.run([
"python", "lantern_scripts/results_visualization.py",
"--gt_dir", gt_exr,
"--gt_render_dir", gt_exr_renders,
"--pandora_dir", renders_lin,
"--pandora_render_dir", lin_renders,
"--out_dir", args.input_dir,
"--name", args.metric_name
], check=True)
+25
View File
@@ -2,12 +2,16 @@ import os
import shutil
import re
import argparse
import subprocess
from pathlib import Path
def copy_files(input_dir):
brackets = os.listdir(input_dir)
target_exr = os.path.join(input_dir, "GT_exr")
target_jpg = os.path.join(input_dir, "GT_jpg")
target_jpg_resize = os.path.join(input_dir, "GT_jpg_resize")
os.makedirs(target_exr, exist_ok=True)
os.makedirs(target_jpg, exist_ok=True)
@@ -36,6 +40,24 @@ def copy_files(input_dir):
shutil.copy2(source_file, destination_file)
print(f"Copied JPG: {source_file} -> {destination_file}")
# Run a simple command
print("Copying target_jpg to target_jpg_resize...")
result = subprocess.run(["cp", "-r", target_jpg, target_jpg_resize], capture_output=True, text=True)
print(result)
print("Resizing target_jpg_resize...")
result = subprocess.run(["find", target_jpg_resize, "-iname", "*.jpg", "-exec", "convert", "{}", "-verbose", "-resize", "3840x1920>", "{}", ";"], capture_output=True, text=True)
print(result)
gt_path = Path(input_dir)
sfm_path = os.path.join(gt_path.parent, 'sfm/images')
for file in os.listdir(target_jpg_resize):
source_file = os.path.join(target_jpg_resize, file)
destination_file = os.path.join(sfm_path, file)
shutil.copy2(source_file, destination_file)
print(f"Copied JPG: {source_file} -> {destination_file}")
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--input_dir', type=str, required=True,
@@ -43,3 +65,6 @@ if __name__ == '__main__':
args = parser.parse_args()
copy_files(args.input_dir)
+75 -7
View File
@@ -2,6 +2,7 @@ from os import makedirs
from html4vision import Col, imagetable
import argparse
import json
import os
if __name__ == "__main__":
@@ -9,13 +10,80 @@ if __name__ == "__main__":
parser.add_argument('--input_dir', type=str, default='/mnt/data/nerfstudio_ds/real_data/coffeeroom_1102H/GT/GT_exr')
args = parser.parse_args()
left_e1_dir = args.input_dir + 'left_e1/'
left_e1_mask_dir = args.input_dir + 'masks/left_e1/'
right_e2_dir = args.input_dir + 'right_e2/'
right_e2_mask_dir = args.input_dir + 'masks/right_e2/'
imgs = [f for f in os.listdir(left_e1_dir) if f.endswith(".png")]
left_e1 = []
left_mask = []
right_e2 = []
right_mask = []
for img in imgs:
left_e1.append( os.path.join(left_e1_dir, img) )
left_mask.append( os.path.join(left_e1_mask_dir, img + '.masked.png') )
right_e2.append( os.path.join(right_e2_dir, img) )
right_mask.append( os.path.join(right_e2_mask_dir, img + '.masked.png') )
cols = [
Col('id1', 'ID'), # 1-based indexing
Col('img', 'Left Pano', args.input_dir + 'left_e1/output_*.png'),
Col('img', 'Left Mask', args.input_dir + 'masks/left_e1/output_*.png.masked.png'),
Col('img', 'Right Pano', args.input_dir + 'right_e2/output_*.png'),
Col('img', 'Right Mask', args.input_dir + 'masks/right_e2/output_*.png.masked.png'),
Col('id1', 'ID'),
Col('text', 'Name', imgs),
Col('img', 'Left Pano', left_e1),
Col('img', 'Left Mask', left_mask),
Col('img', 'Right Pano', right_e2),
Col('img', 'Right Mask', right_mask),
]
# imagetable(cols, 'another-dir/pathrep.html', 'Path Replace Example', pathrep=('/mnt/data', '/home-local2/mokad6.extra.nobkp'))
imagetable(cols, args.input_dir + 'data_index.html', 'NeRF sequence', pathrep=(args.input_dir, './'), imscale=0.1, zebra=True, sticky_header=True, overlay_toggle=True)
imagetable(
cols,
args.input_dir + 'data_index.html',
'NeRF sequence',
pathrep=(args.input_dir, './'),
sort_style='materialize',
imscale=0.1,
zebra=True,
sticky_header=True,
overlay_toggle=True
)
left_sfm_dir = args.input_dir + 'left_sfm/'
left_sfm_mask_dir = args.input_dir + 'masks/left_sfm/'
right_sfm_dir = args.input_dir + 'right_sfm/'
right_sfm_mask_dir = args.input_dir + 'masks/right_sfm/'
imgs = [f for f in os.listdir(left_sfm_dir) if f.endswith(".png")]
left_sfm = []
left_mask = []
right_sfm = []
right_mask = []
for img in imgs:
left_sfm.append( os.path.join(left_sfm_dir, img) )
left_mask.append( os.path.join(left_sfm_mask_dir, img + '.masked.png') )
right_sfm.append( os.path.join(right_sfm_dir, img) )
right_mask.append( os.path.join(right_sfm_mask_dir, img + '.masked.png') )
cols = [
Col('id1', 'ID'),
Col('text', 'Name', imgs),
Col('img', 'Left Pano', left_sfm),
Col('img', 'Left Mask', left_mask),
Col('img', 'Right Pano', right_sfm),
Col('img', 'Right Mask', right_mask),
]
imagetable(
cols,
args.input_dir + 'sfm_index.html',
'SFM sequence',
pathrep=(args.input_dir, './'),
sort_style='materialize',
imscale=0.1,
zebra=True,
sticky_header=True,
overlay_toggle=True
)
+1
View File
@@ -64,6 +64,7 @@ if __name__ == "__main__":
for data_dir in GT_directories:
test_images = []
print("data_dir: ", data_dir)
for path in Path(data_dir).rglob('*.JPG'):
test_images.append(path)
@@ -18,3 +18,5 @@ if __name__ == '__main__':
if args.gt:
subprocess.run(["python", "lantern_scripts/debevec_bracket_merge.py", "--data_dir", args.input_dir + "GT/", "--experiment_location", args.input_dir + "data/"], check=True)
subprocess.run(["python", "lantern_scripts/create_GT_folder.py", "--input_dir", args.input_dir + "GT/"], check=True)
subprocess.run(["python", "lantern_scripts/data_visualization.py", "--input_dir", args.input_dir + "data/"], check=True)
+71
View File
@@ -4,6 +4,9 @@ import subprocess
from pydub import AudioSegment
import numpy as np
from scipy.signal import find_peaks
import json
import re
def extract_audio(video_path, audio_path):
command = [
@@ -15,6 +18,7 @@ def extract_audio(video_path, audio_path):
]
subprocess.run(command)
def detect_hand_claps(audio_path, threshold=0.5, min_distance=10000):
audio = AudioSegment.from_wav(audio_path)
samples = np.array(audio.get_array_of_samples())
@@ -30,6 +34,7 @@ def detect_hand_claps(audio_path, threshold=0.5, min_distance=10000):
return clap_times
def trim_video(video_path, start_time, end_time, output_path):
command = [
"ffmpeg",
@@ -37,10 +42,14 @@ def trim_video(video_path, start_time, end_time, output_path):
"-ss", str(start_time),
"-to", str(end_time),
"-c", "copy",
"-map", "0",
"-map_metadata", "0",
"-movflags", "use_metadata_tags",
output_path
]
subprocess.run(command)
def get_video_length(video_path):
command = [
"ffprobe",
@@ -54,6 +63,48 @@ def get_video_length(video_path):
result = subprocess.run(command, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)
return float(result.stdout.strip())
def extract_imu_data(video_path):
# Run exiftool with -ee to extract per-frame metadata
cmd = ['exiftool', '-ee2', '-n', '-G1', '-a', '-u', video_path]
try:
# Execute the command and capture the output
result = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE) #, text=True, check=True)
output = result.stdout.decode('latin-1')
except subprocess.CalledProcessError as e:
print(f"Error executing ExifTool: {e.stderr}")
return
accel_data = []
gyro_data = []
# Split the output into lines for processing
lines = output.splitlines()
i = 0
while i < len(lines):
line = lines[i]
if 'Accelerometer' in line:
match = re.search(r': ([\-\d\.]+) ([\-\d\.]+) ([\-\d\.]+)', line)
if match:
if i + 1 < len(lines):
if 'Time Stamp' in lines[i + 1]:
time_match = re.search(r': ([\-\d\.]+)', lines[i + 1])
accel_data.append([float(time_match.group(1)), float(match.group(1)), float(match.group(2)), float(match.group(3))])
i += 1
elif 'Magnetic Field' in line:
match = re.search(r': ([\-\d\.]+) ([\-\d\.]+) ([\-\d\.]+)', line)
if match:
if i + 1 < len(lines):
if 'Time Stamp' in lines[i + 1]:
time_match = re.search(r': ([\-\d\.]+)', lines[i + 1])
gyro_data.append([float(time_match.group(1)), float(match.group(1)), float(match.group(2)), float(match.group(3))])
i += 1
else:
i += 1
return accel_data, gyro_data
if __name__ == '__main__':
argparser = argparse.ArgumentParser()
argparser.add_argument("--input_dir", type=str, default="/mnt/data/scene/", help="Directory containing all the 4 videos")
@@ -90,10 +141,30 @@ if __name__ == '__main__':
start_clap_1, end_clap_1 = clap_times_1[0], clap_times_1[-1]
start_clap_2, end_clap_2 = clap_times_2[0], clap_times_2[-1]
accel_data, gyro_data = extract_imu_data(video_path_1)
accel_data_np = np.array(accel_data)
accel_data_np[:, 0] = accel_data_np[:, 0] - accel_data_np[0, 0]
start_index = (np.abs(accel_data_np[:, 0] - start_clap_1)).argmin()
end_index = (np.abs(accel_data_np[:, 0] - end_clap_1)).argmin()
gyro_data_np = np.array(gyro_data)
gyro_data_np[:, 0] = gyro_data_np[:, 0] - gyro_data_np[0, 0]
gyro_start_index = (np.abs(gyro_data_np[:, 0] - start_clap_1)).argmin()
gyro_end_index = (np.abs(gyro_data_np[:, 0] - end_clap_1)).argmin()
imu_data_json = {}
imu_data_json['accel'] = accel_data_np[start_index:end_index, :].tolist()
imu_data_json['gyro'] = gyro_data_np[gyro_start_index:gyro_end_index, :].tolist()
with open(os.path.join(args.input_dir, "imu.json"), "w", encoding="UTF-8") as file:
json.dump(imu_data_json, file, indent=4)
trim_video(video_path_1, start_clap_1, end_clap_1, output_video_path_1)
trim_video(video_path_2, start_clap_2, end_clap_2, output_video_path_2)
if round(get_video_length(output_video_path_1)) != round(get_video_length(output_video_path_2)):
print("clap_times_1: ", clap_times_1, "clap_times_2: ", clap_times_2)
raise ValueError("The videos are not the same lenght.")
os.remove(video_path_1)