JunHo/Cropper
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

module python files

Browse Source
This commit is contained in:
JunHo Kim committed 2021-12-10 19:05:16 +09:00
1 parent ff088e051d
commit 67fdffbe10
5 files changed
+531 -490
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

.gitignore
+5
View File
@@ -1,2 +1,7 @@
__pycache__/*
.vscode/*
backups/*
input/*
logs/*
out/*
test_video/*
README.md
+3 -32
View File
@@ -2,41 +2,12 @@
Convenient Vid cropping tool
## Info
- Window 용 Script
- Window 용 python Script
- 재생은 x3 배속으로 고정
## ivtc Code
< 하나의 파일을 변환하는 코드 >
"ffmpeg -i in.mp4 -vf fieldmatch=mode=pc_n:combmatch=full, bwdif=0:-1:1, decimate -c:v libx264 -crf 0 -preset ultrafast out.mp4"
< 폴더 내에 있는 모든 파일을 변환하는 코드 >
"for %a in ("*.*") do ffmpeg -i "%a" -vf fieldmatch=mode=pc_n:combmatch=full, bwdif=0:-1:1, decimate -c:v libx264 -crf 0 -preset ultrafast ".\out\%~na_ivtc.mp4""
## ToDo
- **Concealing bitstream error** 해결
- ~~타이틀 창 등에 현재 재생 퍼센트 / 프레임 수 Display~~
- 타이틀 한글 깨지는 것
- 처음에 목록 보여주고 시작? (필요한지?)
- Log Ouput 필요
### 포함 해야 할 내용
- 날짜, 시간 정보
- 영상의 정보 (영상 파일명, 해상도, 길이, fps, 포맷)
- ~~Crop 정보 (W, H, X, Y)~~
- 사용자가 Confirm 한 영상 TimeStamp (어디까지 보고 확인했는지에 대한 정보)
- 소요된 Encoding 시간
- ~~Error 로그~~
- ~~File Missing~~
### 로그 파일 포맷
- Txt 파일 포맷
- 영상 파일 명.txt
- 같은 파일에 매 편집 시 누적되는 식
- 각 편집 로그의 구별은 편집 시간정보별로.
- ffmpeg
### 구조적 문제
- Subprocess 기반으로 FFMPEG 동작
- 현재 Encoding 중인 영상이 어느정도 진행 되었는 지 알 수 없음
- 전체 프로그램이 종료되기 전까지는 파일이 생성되지 않음 (subprocess가 계속 돌고있게 됨)
- 엔터를 누를 때 마다 다음 영상으로 넘어가기 때문에, 이전 영상의 진행도를 알기 어려울 듯 하다.
- _생성되는 각 프로세스를 다른 변수명에 Mapping 하여 Log 관리를 각각 해야 하나?_
- Encoding이 종료되면 프로세스를 종료시키는 것 같은 방법이 필요할듯..
- 전체 프로그램이 종료되기 전까지는 파일이 생성되지 않음 (subprocess가 계속 돌고있게 됨)
pillar_detection.py
+42 -458
View File
@@ -1,514 +1,98 @@
import os
from collections import Counter
import cv2
import glob
import re
import subprocess
import argparse
import copy
import datetime
import time
from pillar_detection_utils import *
from pillar_detection_core import *
parser = argparse.ArgumentParser()
# file path options
parser.add_argument("-i", "--input_dir", default="./input", help="Input source vid directory")
parser.add_argument("-o", "--output_dir", default="./out", help="Output vid directory")
# ffmpeg encoding option
parser.add_argument("-c", "--crf", default=20, type=int, help="set crf level")
parser.add_argument("-sp", "--add_sharpening", default=False, action="store_true", help="add ffmpeg sharpening filter")
parser.add_argument("-lb", "--add_letterbox", default=False, action="store_true", help="add letterbox to make 16:9 ratio")
parser.add_argument("-uhd", "--uhd_output", default=False, action="store_true", help="set output resolution to 3840:2160")
parser.add_argument("-lb", "--add_letterbox", default=False, action="store_true", help="add letterbox to make 16:9 ratio")
parser.add_argument("-sp", "--add_sharpening", default=False, action="store_true", help="add ffmpeg sharpening filter")
parser.add_argument("-mul", "--multi_encoding", default=2, type=int, help="when log file exist, encoding mutiple videos by this args")
parser.add_argument("-d", "--debug", default=False, action="store_true", help="for debugging. (short length video output)")
# logfile option
parser.add_argument("-s", "--cropsize_scale", default=1, type=int, help="before use crop size from logfile, scale up the crop size (for reuse when vid rescaled)")
parser.add_argument("-log", "--only_logfile", default=False, action="store_true", help="large vid mode. (only write log file.)")
parser.add_argument("-man", "--manual_mode", default=False, action="store_true", help="manual cropping without using log file (Ignore log files)")
parser.add_argument("-edit", "--edit_mode", default=False, action="store_true", help="get first crop area from log file")
parser.add_argument("-d", "--debug", default=False, action="store_true", help="for debugging. (short length video output)")
args = parser.parse_args()
# option when only write log file (no encoding) : -log
# options when only edit log file that already exists (no encoding) : -log -edit
# Adjust crop info in display layout.
def resizeCropInfo(key, cropPos, frame_width, frame_height):
def _IsOutOfBound(_cropPos):
w, h, x, y = _cropPos
if (w + x) > frame_width or (h + y) > frame_height:
return True
if (x < 0) or (y < 0):
return True
if (w <= 0) or (h <= 0):
return True
w, h, x, y = cropPos
# Left (Expand width)
if key == ord('j') or key == ord('J'):
x -= 2
w += 4
# Right (Shrink width)
elif key == ord('l') or key == ord('L'):
x += 2
w -= 4
# Up (Expand height)
elif key == ord('i') or key == ord('I'):
y -= 2
h += 4
# Down (Shrink height)
elif key == ord('k') or key == ord('K'):
y += 2
h -= 4
# Num 4 (Move left)
elif key == ord('a') or key == ord('A'):
# w -= 1
x -= 1
# Num 6 (Move right)
elif key == ord('d') or key == ord('D'):
# w += 1
x += 1
# Num 8 (Move upward)
elif key == ord('w') or key == ord('W'):
# h -= 1
y -= 1
# Num 2 (Move downward)
elif key == ord('s') or key == ord('S'):
# h += 1
y += 1
_cropPos = w, h, x, y
if not _IsOutOfBound(_cropPos):
cropPos = _cropPos
return cropPos
def getTime():
dt_now = datetime.datetime.now()
dt_str = dt_now.strftime('[%Y-%m-%d %H:%M:%S]')
# print(dt_str)
return dt_str
def getVideDuration(file_path):
ffprobe_command = f"ffprobe -v error -show_entries format=duration -of default=noprint_wrappers=1:nokey=1 \"{file_path}\""
p = subprocess.run(ffprobe_command, shell=False, stdout=subprocess.PIPE)
_hour = 0
_min = 0
_sec = 0
_sec = int(float(p.stdout.decode('utf-8')))
# print(_sec)
if _sec > 60 :
_min = int(_sec // 60)
_sec = _sec - (60 * _min)
# print(_min, _sec)
if _sec > 3600 :
_hour = int(_sec // 60)
_sec = _sec - (60 * _min)
_min = int(_sec // 60)
_sec = _sec - (60 * _min)
# print(_hour, _min, _sec)
return [_hour, _min, _sec]
# for unicode free filename
def imwrite(filename, img, params=None):
try:
ext = os.path.splitext(filename)[1]
result, n = cv2.imencode(ext, img, params)
if result:
with open(filename, mode='w+b') as f:
n.tofile(f)
return True
else:
return False
except Exception as e:
print(e)
return False
def main():
vid_types = ('*.avi', '*.mov', '*.mp4', '*.mxf')
# video encoding opts
input_dir = args.input_dir
crf = args.crf
add_letterbox = args.add_letterbox
add_sharpening = args.add_sharpening
uhd_output = args.uhd_output
# program modes
multi_how_many = args.multi_encoding - 1
scale = int(args.cropsize_scale)
only_logfile = args.only_logfile
manual_mode = args.manual_mode
debug = args.debug
edit_mode = args.edit_mode
input_dir = args.input_dir
logs_dir = input_dir
if scale < 1 :
print(f"[!!! ERROR !!!]\t Invalid scale option [scale : {scale}]")
if args.cropsize_scale < 1 :
print(f"[!!! ERROR !!!]\t Invalid scale option [scale : {args.cropsize_scale}]")
return
elif multi_how_many < 0 :
print(f"[!!! WARRNING !!!]\t \"multi_encoding\" option should be more then 1 [multi_encoding : {multi_how_many}]")
elif args.multi_encoding < 1 :
print(f"[!!! WARRNING !!!]\t \"multi_encoding\" option should be more then 0 [multi_encoding : {args.multi_encoding}]")
return
uhd_output_args = ""
if uhd_output :
uhd_output_args = "-s 3840x2160"
debug_shortOutput = ""
if debug :
debug_shortOutput = "-t 15 -ss 30"
output_dir = args.output_dir
input_list = []
for types in vid_types :
input_list += glob.glob(os.path.join(input_dir, types))
input_list = sorted(input_list)
print("[INFO]\t Input dir :", input_dir)
print(str(input_list))
for vid_index, fpath in enumerate(input_list):
vid_name = os.path.basename(fpath)[:-4]
vid_num = len(input_list)
isSkipped = False
logs_dir = input_dir
logs = ''
log_path = f"{logs_dir}/{vid_name}.txt"
print(f"[INFO]\t [Video num : {vid_index + 1} / {vid_num}]")
print("[INFO]\t File path: ", fpath)
print("[INFO]\t Log dir :", logs_dir)
# init crop configuration class
crop_cfg = VideoCropCfg()
crop_cfg.vid_name = vid_name
# Read log file if exist
if not manual_mode:
if not os.path.exists(f"{logs_dir}/{vid_name}.txt"):
print(f"[INFO]\t Can`t find log file. [{logs_dir}/{vid_name}.txt]")
edit_mode = False
else :
print(f"[INFO]\t Loading log file... [{logs_dir}/{vid_name}.txt]")
f = open(f"{logs_dir}/{vid_name}.txt", "r")
logs = f.read()
cropPos, edit_mode, manual_mode = parseLog(log_path, edit_mode, manual_mode, args.cropsize_scale)
if logs != '':
try :
# Delete "crop=" in list "crop"
allCrops = re.findall("crop=\S+", logs)
cropPos = allCrops[-1].split(":")
cropPos[0] = cropPos[0][5:]
print(f"[INFO]\t Crop area from log file: [w : {cropPos[0]}, h : {cropPos[1]}, x : {cropPos[2]}, y : {cropPos[3]}]")
cropPos = [int(x * scale) for x in cropPos]
if scale > 1 :
print(f"[INFO]\t Scale up crop area: [w : {cropPos[0]}, h : {cropPos[1]}, x : {cropPos[2]}, y : {cropPos[3]}]")
except:
print(f"[!!! WARNNIG !!!]\t Can`t read log file. [{logs_dir}/{vid_name}.txt]")
else :
print(f"[!!! WARNNIG !!!]\t Empty log file. [{logs_dir}/{vid_name}.txt]")
if (logs == '') or manual_mode or edit_mode:
# TODO : Korean title text bug fix
# vid_name = vid_name.encode('utf-8').decode('cp949')
print(f"[INFO]\t [Video num : {vid_index + 1} / {vid_num}]")
print("[INFO]\t File to detect crop area: ", fpath)
if manual_mode or edit_mode:
# Capture vid, Get Vid info
crop_cfg.getVidInfo(fpath)
if edit_mode:
print("[INFO]\t (Edit mode) Read crop area from log file...")
print(f"[INFO]\t Crop area from log file: [w : {cropPos[0]}, h : {cropPos[1]}, x : {cropPos[2]}, y : {cropPos[3]}]")
else :
print("[INFO]\t Detecting crop area...")
p = subprocess.Popen(["ffmpeg", "-ss", "10", "-i", fpath, "-vf", "cropdetect=limit=38",
"-vframes", "3000", "-f", "null", "out.null"], stdout=subprocess.PIPE, stderr=subprocess.PIPE)
infos = p.stderr.read().decode('utf-8')
allCrops = re.findall("crop=\S+", infos)
mostCommonCrop = Counter(allCrops).most_common(1)
if len(mostCommonCrop) > 0 :
crop = mostCommonCrop[0][0]
else :
print(f"[!!! ERROR !!!]\t Can`t read input vid. Skip to next vid \t ({fpath})")
continue
# Delete "crop=" in list "crop"
cropPos = crop.split(":")
cropPos[0] = cropPos[0][5:]
print(f"[INFO]\t Most common detected crop area: [w : {cropPos[0]}, h : {cropPos[1]}, x : {cropPos[2]}, y : {cropPos[3]}]")
cropPos = [int(x) for x in cropPos]
# Capture vid, Get Vid info
capture = cv2.VideoCapture(fpath)
frame_width = capture.get(cv2.CAP_PROP_FRAME_WIDTH)
frame_height = capture.get(cv2.CAP_PROP_FRAME_HEIGHT)
fps = capture.get(cv2.CAP_PROP_FPS)
frame_pos = 0
frame_count = 0
# Crop Area even num check
if (cropPos[0] % 2) != 0 :
if (cropPos[0] + 1) >= frame_width:
cropPos[0] -= 1
else:
cropPos[0] += 1
if (cropPos[1] % 2) != 0 :
if (cropPos[1] + 1) >= frame_width:
cropPos[1] -= 1
else:
cropPos[1] += 1
# Crop Area Multiple of 4 check
if (cropPos[0] % 4) != 0 :
if (cropPos[0] + 2) >= frame_width:
cropPos[0] -= 2
else:
cropPos[0] += 2
if (cropPos[1] % 4) != 0 :
if (cropPos[1] + 2) >= frame_width:
cropPos[1] -= 2
else:
cropPos[1] += 2
key = None
isPaused = False
cleanFrame = None
_boxedFrame = None
frame = None
cropPos = autoDetectCropArea(fpath)
# Crop Area validation check (width, height must be even number)
cropPos = adjustCropArea(cropPos)
cv2.namedWindow(f'{vid_name}', cv2.WINDOW_NORMAL | cv2.WINDOW_KEEPRATIO)
# Display crop size control UI
while True:
key = cv2.waitKeyEx(10)
# Close button(X button) event
if cv2.getWindowProperty(f'{vid_name}', cv2.WND_PROP_VISIBLE) < 1 :
return
if key > 0 :
# 13 == Enter key ASCII Code
if key == 13:
break
# Pause
elif key == ord('p') or key == ord('P'):
if isPaused :
isPaused = False
else :
isPaused = True
# frame = copy.deepcopy(cleanFrame)
# Replay
elif key == ord('r') or key == ord('R'):
if not isPaused:
capture = cv2.VideoCapture(fpath)
continue
# Move Forward (5s)
elif key == ord('e') or key == ord('E'):
if not isPaused:
capture.set(cv2.CAP_PROP_POS_FRAMES, frame_pos + (fps * 5))
# Move Backward (5s)
elif key == ord('q') or key == ord('Q'):
if not isPaused:
capture.set(cv2.CAP_PROP_POS_FRAMES, frame_pos - (fps * 5))
# Display key guide
elif key == ord('h') or key == ord('H'):
print("[INFO]\t ### KEY GUIDE ##############################################################################################################################################################")
print("[INFO]\t # (MOVE : wasd), (RESIZE : ijkl), (FOR / BACKWARD : q, e), (RESET : r), (PAUSE : p), (SKIP : n), (Preview : v), (Save current frame : b)(CONFIRM : enter) #")
print("[INFO]\t ############################################################################################################################################################################ ")
# Skip this Vid
elif key == ord('n') or key == ord('N'):
print(f"[INFO]\t Skip this video")
isSkipped = True
break
# Preview this crop area on new window
elif key == ord('v') or key == ord('V'):
_x = cropPos[2]
_y = cropPos[3]
_w = cropPos[0]
_h = cropPos[1]
_frame = copy.deepcopy(cleanFrame)
cv2.namedWindow(f'{vid_name}_current_area', cv2.WINDOW_KEEPRATIO)
cv2.waitKey(10)
cv2.imshow(f'{vid_name}_current_area', _frame[_y: (_y + _h), _x: (_x + _w)])
# Capture, download this crop area in png file
elif key == ord('b') or key == ord('B'):
_x = cropPos[2]
_y = cropPos[3]
_w = cropPos[0]
_h = cropPos[1]
_frame = copy.deepcopy(cleanFrame)
_vid_name = vid_name.replace(" ","_")
image_dir = "./capture_image"
path = os.path.join(image_dir, _vid_name)
if not os.path.exists(path) :
os.makedirs(path)
print(f'[INFO]\t {path} created')
image_name = f'{_vid_name}_({frame_pos}-{frame_count})_before.png'
_path = os.path.join(path, image_name)
imwrite(_path, _boxedFrame)
print(f"\n[INFO]\t Saved current frame : {_path}")
image_name = f'{_vid_name}_({frame_pos}-{frame_count})_after.png'
_path = os.path.join(path, image_name)
imwrite(_path, _frame[_y: (_y + _h), _x: (_x + _w)])
print(f"[INFO]\t Saved current frame in crop area : {_path}")
else:
# Crop Area control
cropPos = resizeCropInfo(key, cropPos, frame_width, frame_height)
if not isPaused:
frame_pos = int(capture.get(cv2.CAP_PROP_POS_FRAMES))
frame_count = int(capture.get(cv2.CAP_PROP_FRAME_COUNT))
if(frame_pos >= frame_count):
retCode, cropPos = video_croppping(crop_cfg, cropPos)
if retCode == 1:
break
# When paused, hold current frame untill unpause.
if isPaused:
# deep copy current frame, use that frame for draw rectangle.
# cleanFrame = copy.deepcopy(frame)
_boxedFrame = copy.deepcopy(cleanFrame)
cv2.rectangle(_boxedFrame, (cropPos[2], cropPos[3]), (cropPos[2] + cropPos[0],
cropPos[3] + cropPos[1]), (0, 0, 255), thickness=2, lineType=cv2.LINE_8)
cv2.imshow(f'{vid_name}', _boxedFrame)
else :
ret, frame = capture.read()
# deep copy current frame for clean preview.
cleanFrame = copy.deepcopy(frame)
cv2.rectangle(frame, (cropPos[2], cropPos[3]), (cropPos[2] + cropPos[0],
cropPos[3] + cropPos[1]), (0, 0, 255), thickness=2, lineType=cv2.LINE_8)
cv2.imshow(f'{vid_name}', frame)
_boxedFrame = copy.deepcopy(frame)
elif retCode == 2:
crop_cfg.getVidInfo(fpath)
continue
print("[INFO]\t Current crop area : ", end="")
print(cropPos, end="")
print(f"\t Frame Count ({frame_pos}/{frame_count})\t[PRESS \"h\" TO VIEW KEY GUIDE]", end="\r")
print("")
capture.release()
crop_cfg.vidRelease()
cv2.destroyAllWindows()
# Debugging
# cropPos[0] = 678
# cropPos[1] = 402
# cropPos[2] = 20
# cropPos[3] = 38
# pad_arg = ""
dt_str = getTime()
if not isSkipped:
if not os.path.exists(f"{logs_dir}"):
print(f"[INFO]\t Make log directory [{logs_dir}/{vid_name}.txt]")
os.makedirs(f"{logs_dir}")
if not os.path.exists(f"{logs_dir}/{vid_name}.txt"):
print(f"[INFO]\t Save log file in [{logs_dir}/{vid_name}.txt]")
f = open(f"{logs_dir}/{vid_name}.txt", 'w')
else :
print(f"[INFO]\t Add line in log file [{logs_dir}/{vid_name}.txt]")
f = open(f"{logs_dir}/{vid_name}.txt", 'a')
f.write(f"{dt_str}\tcrop={cropPos[0]}:{cropPos[1]}:{cropPos[2]}:{cropPos[3]}\n")
f.close
writeLog(logs_dir, log_path, cropPos)
# Set ffmpeg commands, actual ffmpeg encoding
if not (only_logfile or isSkipped):
crop_arg = f"crop={cropPos[0]}:{cropPos[1]}:{cropPos[2]}:{cropPos[3]}"
print(f"[INFO]\t [CropInfo] (W : {cropPos[0]}), (H : {cropPos[1]}), (X : {cropPos[2]}), (Y : {cropPos[3]})")
crop_arg = f"crop={cropPos[0]}:{cropPos[1]}:{cropPos[2]}:{cropPos[3]}"
current_crop_ratio = cropPos[0] / cropPos[1]
scale_arg = ""
pad_arg = ""
"""
Set padding(letter box) argument
variable "offset" for 16:9 output display ratio
pad_arg syntax:
pad= {width after padding}:{height after padding}:{x offset of pad}:{y offset of pad}
"""
if add_letterbox :
# Always target 16:9 display ratio
target_ratio = 16 / 9
print("[INFO]\t Target ratio : %.7f" %target_ratio)
# when current ratio is more flat
if target_ratio < current_crop_ratio:
# offset : simple equation (widht / (height + offset) = 16/9)
offset = (9 * cropPos[0] - 16 * cropPos[1]) / 16
tmp_height = cropPos[1] + offset
tmp_height = int(tmp_height)
if (tmp_height % 4) != 0:
tmp_height -= (tmp_height % 4)
pad_arg = f",pad={cropPos[0]}:{tmp_height}:(ow-iw)/2:(ih-oh)/2"
scale_arg = f",scale=w={cropPos[0]}:h={tmp_height}"
print(f"[INFO]\t [PadInfo] (W : {cropPos[0]}), (H : {tmp_height})")
else:
offset = (16 * cropPos[1] - 9 * cropPos[0]) / 9
tmp_width = cropPos[0] + offset
tmp_width = int(tmp_width)
if (tmp_width % 4) != 0:
tmp_width -= (tmp_width % 4)
pad_arg = f",pad={tmp_width}:{cropPos[1]}:(ow-iw)/2:(ih-oh)/2"
scale_arg = f",scale=w={tmp_width}:h={cropPos[1]}"
print(f"[INFO]\t [PadInfo] (W : {tmp_width}), (H : {cropPos[1]})")
if not os.path.exists(output_dir) :
os.mkdir(output_dir)
print(f'[INFO]\t folder created : {output_dir}')
output_path = os.path.join(output_dir, f"{vid_name}_box.mov")
## ffmpeg command setting
if add_sharpening :
sharpen_arg = ',unsharp=7:7:1.5:7:7:1.5'
else :
sharpen_arg = ''
'''
ffmpeg arguments
-stats : print only encoding progress status (ignore loglevel argument)
-loglevel : make ffmpeg stdout more verbose or not
warning : only warning, error
error : only errors
info : defualt
'''
loglevel = 'error'
ffmpeg_command = f"ffmpeg {debug_shortOutput} -i \"{fpath}\" -vf \"{crop_arg}{scale_arg}{pad_arg}{sharpen_arg}\"\
-y -pix_fmt yuv420p -c:v libx264 -crf {crf} -c:a copy -preset medium -loglevel {loglevel}\
{uhd_output_args} \"{output_path}\""
print(ffmpeg_command)
print("[INFO]\t Encoding \"%s\" ..." % vid_name)
print("[INFO]\t Video CRF = %d" % int(crf))
# duration_list[0] : hour, [1] : min, [2] : sec
duration_list = getVideDuration(fpath)
print(f"[INFO]\t Video Duration : {duration_list[0]:02d}:{duration_list[1]:02d}:{duration_list[2]:02d}\
Current Time : {getTime()}")
st_time = time.time()
if (logs == '') or manual_mode:
# Manual Mode (always encoding one by one)
# Hold process when encoding last video
if (vid_index + 1) == vid_num :
subprocess.Popen(ffmpeg_command + ' -stats', shell=False).wait()
else :
subprocess.Popen(ffmpeg_command, shell=False)
else:
# Log file mode (encoding simultaneously n videos)
if multi_how_many == 0:
p = subprocess.Popen(ffmpeg_command + ' -stats', shell=False).wait()
# [21.12.08] progress status coding..
# p = subprocess.Popen(ffmpeg_command, shell=False, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
# while p.poll() is None:
# # print('now encoding..', end='\r')
# infos = p.stderr.read().decode('utf-8')
# print('h',infos)
# infos = p.stdout.read().decode('utf-8')
# print('t',infos)
# continue
elif ((vid_index % multi_how_many == 0) and (vid_index != 0)) or ((vid_index + 1) == vid_num):
subprocess.Popen(ffmpeg_command + ' -stats', shell=False).wait()
else :
subprocess.Popen(ffmpeg_command, shell=False)
if not (args.only_logfile or isSkipped):
ffmpeg_encoding(args, vid_index, vid_num, cropPos, vid_name, fpath)
if __name__ == "__main__":
main()
pillar_detection_core.py
+269
View File
@@ -0,0 +1,269 @@
from pillar_detection_utils import *
class VideoCropCfg():
def __init__(self):
self.vid_name = ''
self.frame_pos = 0
self.frame_count = 0
self.captured_vid = None
self.isPaused = False
self.isSkipped = False
self.cleanFrame = None
self.boxedFrame = None
self.frame = None
self.frame_width = 0
self.frame_height = 0
self.fps = 0
def getVidInfo(self, fpath):
self.captured_vid = cv2.VideoCapture(fpath)
self.frame_width = self.captured_vid.get(cv2.CAP_PROP_FRAME_WIDTH)
self.frame_height = self.captured_vid.get(cv2.CAP_PROP_FRAME_HEIGHT)
self.fps = self.captured_vid.get(cv2.CAP_PROP_FPS)
def vidRelease(self):
self.captured_vid.release()
def video_croppping(crop_cfg, cropPos):
'''
retCode value mean
0 : keep playing
1 : done with cropping (to next vid)
2 : replay
'''
retCode = 0
# search_distance unit : second
search_distance = 5
# play_speed unit : millisecond
play_speed = 10
key = cv2.waitKeyEx(play_speed)
# Close button(X button) event
if cv2.getWindowProperty(f'{crop_cfg.vid_name}', cv2.WND_PROP_VISIBLE) < 1 :
retCode = 1
return retCode, cropPos
if key > 0 :
# 13 == Enter key ASCII Code
if key == 13:
retCode = 1
# Pause
elif key == ord('p') or key == ord('P'):
if crop_cfg.isPaused :
crop_cfg.isPaused = False
else :
crop_cfg.isPaused = True
# Replay
elif key == ord('r') or key == ord('R'):
if not crop_cfg.isPaused:
retCode = 2
# Move Forward
elif key == ord('e') or key == ord('E'):
if not crop_cfg.isPaused:
crop_cfg.captured_vid.set(cv2.CAP_PROP_POS_FRAMES, crop_cfg.frame_pos + (crop_cfg.fps * search_distance))
# Move Backward
elif key == ord('q') or key == ord('Q'):
if not crop_cfg.isPaused:
crop_cfg.captured_vid.set(cv2.CAP_PROP_POS_FRAMES, crop_cfg.frame_pos - (crop_cfg.fps * search_distance))
# Display key guide
elif key == ord('h') or key == ord('H'):
print("[INFO]\t ### KEY GUIDE ##############################################################################################################################################################")
print("[INFO]\t # (MOVE : wasd), (RESIZE : ijkl), (FOR / BACKWARD : q, e), (RESET : r), (PAUSE : p), (SKIP : n), (Preview : v), (Save current frame : b)(CONFIRM : enter) #")
print("[INFO]\t ############################################################################################################################################################################ ")
# Skip this Vid
elif key == ord('n') or key == ord('N'):
print(f"[INFO]\t Skip this video")
crop_cfg.isSkipped = True
retCode = 1
# Preview this crop area on new window
elif key == ord('v') or key == ord('V'):
_x = cropPos[2]
_y = cropPos[3]
_w = cropPos[0]
_h = cropPos[1]
_frame = copy.deepcopy(crop_cfg.cleanFrame)
cv2.namedWindow(f'{crop_cfg.vid_name}_current_area', cv2.WINDOW_KEEPRATIO)
cv2.waitKey(10)
cv2.imshow(f'{crop_cfg.vid_name}_current_area', _frame[_y: (_y + _h), _x: (_x + _w)])
# captured_vid, download this crop area in png file
elif key == ord('b') or key == ord('B'):
_x = cropPos[2]
_y = cropPos[3]
_w = cropPos[0]
_h = cropPos[1]
_frame = copy.deepcopy(crop_cfg.cleanFrame)
_vid_name = crop_cfg.vid_name.replace(" ","_")
image_dir = "./capture_image"
path = os.path.join(image_dir, _vid_name)
if not os.path.exists(path) :
os.makedirs(path)
print(f'[INFO]\t {path} created')
image_name = f'{_vid_name}_({crop_cfg.frame_pos}-{crop_cfg.frame_count})_before.png'
_path = os.path.join(path, image_name)
imwrite(_path, crop_cfg.boxedFrame)
print(f"\n[INFO]\t Saved current frame : {_path}")
image_name = f'{_vid_name}_({crop_cfg.frame_pos}-{crop_cfg.frame_count})_after.png'
_path = os.path.join(path, image_name)
imwrite(_path, _frame[_y: (_y + _h), _x: (_x + _w)])
print(f"[INFO]\t Saved current frame in crop area : {_path}")
else:
# Crop Area control
cropPos = resizeCropInfo(key, cropPos, crop_cfg.frame_width, crop_cfg.frame_height)
if not crop_cfg.isPaused:
crop_cfg.frame_pos = int(crop_cfg.captured_vid.get(cv2.CAP_PROP_POS_FRAMES))
crop_cfg.frame_count = int(crop_cfg.captured_vid.get(cv2.CAP_PROP_FRAME_COUNT))
if(crop_cfg.frame_pos >= crop_cfg.frame_count):
retCode = 1
# When paused, hold current frame untill unpause.
if crop_cfg.isPaused:
# deep copy current frame, use that frame for draw rectangle.
crop_cfg.boxedFrame = copy.deepcopy(crop_cfg.cleanFrame)
cv2.rectangle(crop_cfg.boxedFrame, (cropPos[2], cropPos[3]), (cropPos[2] + cropPos[0],
cropPos[3] + cropPos[1]), (0, 0, 255), thickness=2, lineType=cv2.LINE_8)
cv2.imshow(f'{crop_cfg.vid_name}', crop_cfg.boxedFrame)
else :
_, crop_cfg.frame = crop_cfg.captured_vid.read()
# deep copy current frame for clean preview.
crop_cfg.cleanFrame = copy.deepcopy(crop_cfg.frame)
cv2.rectangle(crop_cfg.frame, (cropPos[2], cropPos[3]), (cropPos[2] + cropPos[0],
cropPos[3] + cropPos[1]), (0, 0, 255), thickness=2, lineType=cv2.LINE_8)
cv2.imshow(f'{crop_cfg.vid_name}', crop_cfg.frame)
crop_cfg.boxedFrame = copy.deepcopy(crop_cfg.frame)
print("[INFO]\t Current crop area : ", end="")
print(cropPos, end="")
print(f"\t Frame Count ({crop_cfg.frame_pos}/{crop_cfg.frame_count})\t[PRESS \"h\" TO VIEW KEY GUIDE]", end="\r")
return retCode, cropPos
def ffmpeg_encoding(args, vid_index, vid_num, cropPos, vid_name, fpath):
uhd_output = args.uhd_output
debug = args.debug
add_letterbox = args.add_letterbox
output_dir = args.output_dir
add_sharpening = args.add_sharpening
manual_mode = args.manual_mode
crf = args.crf
uhd_output_args = ""
if uhd_output :
uhd_output_args = "-s 3840x2160"
debug_shortOutput = ""
if debug :
debug_shortOutput = "-t 15 -ss 30"
crop_arg = f"crop={cropPos[0]}:{cropPos[1]}:{cropPos[2]}:{cropPos[3]}"
print(f"[INFO]\t [CropInfo] (W : {cropPos[0]}), (H : {cropPos[1]}), (X : {cropPos[2]}), (Y : {cropPos[3]})")
crop_arg = f"crop={cropPos[0]}:{cropPos[1]}:{cropPos[2]}:{cropPos[3]}"
current_crop_ratio = cropPos[0] / cropPos[1]
scale_arg = ""
pad_arg = ""
"""
Set padding(letter box) argument
variable "offset" for 16:9 output display ratio
pad_arg syntax:
pad= {width after padding}:{height after padding}:{x offset of pad}:{y offset of pad}
"""
if add_letterbox :
# Always target 16:9 display ratio
target_ratio = 16 / 9
print("[INFO]\t Target ratio : %.7f" %target_ratio)
# when current ratio is more flat
if target_ratio < current_crop_ratio:
# offset : simple equation (widht / (height + offset) = 16/9)
offset = (9 * cropPos[0] - 16 * cropPos[1]) / 16
tmp_height = cropPos[1] + offset
tmp_height = int(tmp_height)
if (tmp_height % 4) != 0:
tmp_height -= (tmp_height % 4)
pad_arg = f",pad={cropPos[0]}:{tmp_height}:(ow-iw)/2:(ih-oh)/2"
scale_arg = f",scale=w={cropPos[0]}:h={tmp_height}"
print(f"[INFO]\t [PadInfo] (W : {cropPos[0]}), (H : {tmp_height})")
else:
offset = (16 * cropPos[1] - 9 * cropPos[0]) / 9
tmp_width = cropPos[0] + offset
tmp_width = int(tmp_width)
if (tmp_width % 4) != 0:
tmp_width -= (tmp_width % 4)
pad_arg = f",pad={tmp_width}:{cropPos[1]}:(ow-iw)/2:(ih-oh)/2"
scale_arg = f",scale=w={tmp_width}:h={cropPos[1]}"
print(f"[INFO]\t [PadInfo] (W : {tmp_width}), (H : {cropPos[1]})")
if not os.path.exists(output_dir) :
os.mkdir(output_dir)
print(f'[INFO]\t folder created : {output_dir}')
output_path = os.path.join(output_dir, f"{vid_name}_box.mov")
## ffmpeg command setting
if add_sharpening :
sharpen_arg = ',unsharp=7:7:1.5:7:7:1.5'
else :
sharpen_arg = ''
'''
ffmpeg arguments
-stats : print only encoding progress status (ignore loglevel argument)
-loglevel : make ffmpeg stdout more verbose or not
warning : only warning, error
error : only errors
fatal : only fatal (crash) errors
info : defualt
'''
loglevel = 'fatal'
ffmpeg_command = f"ffmpeg {debug_shortOutput} -i \"{fpath}\" -vf \"{crop_arg}{scale_arg}{pad_arg}{sharpen_arg}\"\
-y -pix_fmt yuv420p -c:v libx264 -crf {crf} -c:a copy -preset medium -loglevel {loglevel}\
{uhd_output_args} \"{output_path}\""
print(ffmpeg_command)
print("[INFO]\t Encoding \"%s\" ..." % vid_name)
print("[INFO]\t Video CRF = %d" % int(crf))
# duration_list[0] : hour, [1] : min, [2] : sec
duration_list = getVideDuration(fpath)
print(f"[INFO]\t Video Duration : {duration_list[0]:02d}:{duration_list[1]:02d}:{duration_list[2]:02d}\
Current Time : {getTime()}")
# st_time = time.time()
vid_index += 1
if manual_mode:
# Manual Mode (always encoding one by one)
# Hold process when encoding last video
if vid_index == vid_num :
subprocess.Popen(ffmpeg_command + ' -stats', shell=False).wait()
else :
subprocess.Popen(ffmpeg_command, shell=False)
else:
# Log file mode (encoding simultaneously n videos)
if args.multi_encoding == 0:
p = subprocess.Popen(ffmpeg_command + ' -stats', shell=False).wait()
# [21.12.08] progress status coding..
# p = subprocess.Popen(ffmpeg_command, shell=False, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
# while p.poll() is None:
# # print('now encoding..', end='\r')
# infos = p.stderr.read().decode('utf-8')
# print('h',infos)
# infos = p.stdout.read().decode('utf-8')
# print('t',infos)
# continue
elif (vid_index % args.multi_encoding == 0) or (vid_index == vid_num):
subprocess.Popen(ffmpeg_command + ' -stats', shell=False).wait()
else :
subprocess.Popen(ffmpeg_command, shell=False)
pillar_detection_utils.py
+212
View File
@@ -0,0 +1,212 @@
import os
from collections import Counter
import cv2
import glob
import re
import subprocess
import argparse
import copy
import datetime
import time
def resizeCropInfo(key, cropPos, frame_width, frame_height):
# Adjust crop info in display layout.
def _IsOutOfBound(_cropPos):
w, h, x, y = _cropPos
if (w + x) > frame_width or (h + y) > frame_height:
return True
if (x < 0) or (y < 0):
return True
if (w <= 0) or (h <= 0):
return True
w, h, x, y = cropPos
# Left (Expand width)
if key == ord('j') or key == ord('J'):
x -= 2
w += 4
# Right (Shrink width)
elif key == ord('l') or key == ord('L'):
x += 2
w -= 4
# Up (Expand height)
elif key == ord('i') or key == ord('I'):
y -= 2
h += 4
# Down (Shrink height)
elif key == ord('k') or key == ord('K'):
y += 2
h -= 4
# Num 4 (Move left)
elif key == ord('a') or key == ord('A'):
# w -= 1
x -= 1
# Num 6 (Move right)
elif key == ord('d') or key == ord('D'):
# w += 1
x += 1
# Num 8 (Move upward)
elif key == ord('w') or key == ord('W'):
# h -= 1
y -= 1
# Num 2 (Move downward)
elif key == ord('s') or key == ord('S'):
# h += 1
y += 1
_cropPos = w, h, x, y
if not _IsOutOfBound(_cropPos):
cropPos = _cropPos
return cropPos
def getTime():
dt_now = datetime.datetime.now()
dt_str = dt_now.strftime('[%Y-%m-%d %H:%M:%S]')
# print(dt_str)
return dt_str
def getVideDuration(file_path):
ffprobe_command = f"ffprobe -v error -show_entries format=duration -of default=noprint_wrappers=1:nokey=1 \"{file_path}\""
p = subprocess.run(ffprobe_command, shell=False, stdout=subprocess.PIPE)
_hour = 0
_min = 0
_sec = 0
_sec = int(float(p.stdout.decode('utf-8')))
# print(_sec)
if _sec > 60 :
_min = int(_sec // 60)
_sec = _sec - (60 * _min)
# print(_min, _sec)
if _sec > 3600 :
_hour = int(_sec // 60)
_sec = _sec - (60 * _min)
_min = int(_sec // 60)
_sec = _sec - (60 * _min)
# print(_hour, _min, _sec)
return [_hour, _min, _sec]
# for unicode free filename
def imwrite(filename, img, params=None):
try:
ext = os.path.splitext(filename)[1]
result, n = cv2.imencode(ext, img, params)
if result:
with open(filename, mode='w+b') as f:
n.tofile(f)
return True
else:
return False
except Exception as e:
print(e)
return False
def parseLog(log_path, edit_mode, manual_mode, scale):
cropPos = []
if not os.path.exists(log_path):
print(f"[INFO]\t Can`t find log file. [{log_path}]")
manual_mode = True
edit_mode = False
else :
print(f"[INFO]\t Loading log file... [{log_path}]")
try :
f = open(log_path, "r")
logs = f.read()
# if log file is not empty
if logs != '':
# Delete "crop=" in list "crop"
allCrops = re.findall("crop=\S+", logs)
cropPos = allCrops[-1].split(":")
cropPos[0] = cropPos[0][5:]
print(f"[INFO]\t Crop area from log file: [w : {cropPos[0]}, h : {cropPos[1]}, x : {cropPos[2]}, y : {cropPos[3]}]")
cropPos = [int(x * scale) for x in cropPos]
if scale > 1 :
print(f"[INFO]\t Scale up crop area: [w : {cropPos[0]}, h : {cropPos[1]}, x : {cropPos[2]}, y : {cropPos[3]}]")
else :
print(f"[!!! WARNNIG !!!]\t Empty log file. [{log_path}]")
manual_mode = True
edit_mode = False
except Exception as e:
print(f"[!!! WARNNIG !!!]\t Can`t read log file. [{log_path}]")
print(e)
manual_mode = True
edit_mode = False
return cropPos, edit_mode, manual_mode
def writeLog(logs_dir, log_path, cropPos):
if not os.path.exists(f"{logs_dir}"):
print(f"[INFO]\t Make log directory [{log_path}]")
os.makedirs(f"{logs_dir}")
if not os.path.exists(f"{log_path}"):
print(f"[INFO]\t Save log file in [{log_path}]")
f = open(f"{log_path}", 'w')
else :
print(f"[INFO]\t Add line in log file [{log_path}]")
f = open(f"{log_path}", 'a')
dt_str = getTime()
f.write(f"{dt_str}\tcrop={cropPos[0]}:{cropPos[1]}:{cropPos[2]}:{cropPos[3]}\n")
f.close
def autoDetectCropArea(fpath):
cropPos = []
print("[INFO]\t Detecting crop area...")
p = subprocess.Popen(["ffmpeg", "-ss", "10", "-i", fpath, "-vf", "cropdetect=limit=38",
"-vframes", "3000", "-f", "null", "out.null"], stdout=subprocess.PIPE, stderr=subprocess.PIPE)
infos = p.stderr.read().decode('utf-8')
allCrops = re.findall("crop=\S+", infos)
mostCommonCrop = Counter(allCrops).most_common(1)
if len(mostCommonCrop) > 0 :
crop = mostCommonCrop[0][0]
else :
print(f"[!!! ERROR !!!]\t Can`t read input vid. Skip to next vid \t ({fpath})")
return cropPos
# Delete "crop=" in list "crop"
cropPos = crop.split(":")
cropPos[0] = cropPos[0][5:]
print(f"[INFO]\t Most common detected crop area: [w : {cropPos[0]}, h : {cropPos[1]}, x : {cropPos[2]}, y : {cropPos[3]}]")
cropPos = [int(x) for x in cropPos]
return cropPos
def adjustCropArea(cropPos):
# Crop Area even num check
if (cropPos[0] % 2) != 0 :
if (cropPos[0] + 1) >= frame_width:
cropPos[0] -= 1
else:
cropPos[0] += 1
if (cropPos[1] % 2) != 0 :
if (cropPos[1] + 1) >= frame_width:
cropPos[1] -= 1
else:
cropPos[1] += 1
# Crop Area Multiple of 4 check
if (cropPos[0] % 4) != 0 :
if (cropPos[0] + 2) >= frame_width:
cropPos[0] -= 2
else:
cropPos[0] += 2
if (cropPos[1] % 4) != 0 :
if (cropPos[1] + 2) >= frame_width:
cropPos[1] -= 2
else:
cropPos[1] += 2
return cropPos